Metabolic Mechanisms for the Evolution of Stable Symbiosis

Metabolic mechanisms for the evolution of stable symbiosis

Megan Elisabeth Stig Sørensen

A thesis submitted for the degree of Doctor of Philosophy

University of Sheffield Department of Animal and Plant Sciences

September 2019

Abstract

Endosymbiosis involves the merger of once independent organisms; this evolutionary transition has defined the evolutionary history of eukaryotes and continues to underpin the function of a wide range of ecosystems. Endosymbioses are evolutionarily dynamic because the inherent conflict between the self-interest of the partners make the breakdown of the interaction ever-likely and this is exacerbated by the environmental context dependence of the benefits of symbiosis. This necessitates selection for partner switching, which can reshuffle the genetic identities of symbiotic partnerships and so rescue symbioses from cheater-induced extinction and enable rapid adaptation to environmental change. However, the mechanisms of partner-specificity, that underlie the potential for partner switching, are unknown. Here I report the metabolic mechanisms that control partner specificity within the tractable microbial photosymbiosis between Paramecium bursaria and Chlorella. I have found that metabolic function, and not genetic identity, enables partner-switching, but that genetic variation plays an important role in maintaining variation in symbiotic phenotype. In addition, I observed that symbiont stress-responses played an important role in partner specificity, and that alleviating symbiont stress responses may be an important strategy of generalist host genotypes. Furthermore, I have used experimental evolution to show that a novel, initially non-beneficial association can rapidly evolve to become a beneficial symbiosis. These results demonstrate that partner integration is defined by metabolic compatibility and that initially maladapted host- symbiont pairings can rapidly evolve to overcome their lack of co-adaptation through alterations to metabolism and symbiont regulation. Understanding the process of novel partner integration and partner switching is crucial if we are to understand how new symbioses originate and stabilise. Moreover, mechanistic knowledge of partner switching is required to mitigate the breakdown of symbioses performing important ecosystem functions driven by environmental change, such as in coral reefs.

List of Contents

Abstract 3 List of Figures 6 List of Tables 8 Acknowledgements 9 Declaration 10

Chapter 1 – Introduction 11 1.1 The Organelles 12 1.2 Secondary Endosymbioses 14 1.3 The parasitism-mutualism continuum 15 1.4 Evolution of partner dependency 17 1.5 Conflict avoidance 19 1.6 Ecology and Physiology of the P. bursaria – Chlorella endosymbiosis 20 1.7 Genetics of the P. bursaria – Chlorella endosymbiosis 26 1.8 Thesis Outline 29

Chapter 2 – Comparison of independent evolutionary origins reveals both convergence and divergence in the metabolic mechanisms of symbiosis 31 2.1 Introduction 31 2.2 Materials and Methods 33 2.3 Results 39 2.4 Discussion 51 2.5 Supplementary Figures 56 2.6 Supplementary Results 61

Chapter 3 – Light-dependent stress-responses underlie host-symbiont genotypic specificity in a photosymbiosis 62 3.1 Introduction 62 3.2 Materials & Methods 65 3.3 Results 68 3.4 Discussion 81 3.5 Supplementary Figures 87 3.6 Supplementary Tables 91

Chapter 4 – A novel host-symbiont interaction can rapidly evolve to become a beneficial symbiosis 93 4.1 Introduction 93 4.2 Materials and Methods 95 4.3 Results 98 4.4 Discussion 105 4.5 Supplementary Tables 110

Chapter 5 – Discussion 114 5.1 Stress and symbiosis 115 5.2 Partner Switching 116 5.3 Rapid evolution enables the establishment of symbiosis 118 5.4 Applications of endosymbiosis research 119 5.5 Future-directions 120 5.6 In conclusion 121

Appendix A – The review paper linked to Chapter 1 123 Appendix B – Statistical outputs for Chapter 2 131 Appendix C – Statistical outputs for Chapter 3 134 Appendix D – Statistical outputs for Chapter 4 137

Bibliography 139

List of Figures

1.1 Diagrammatic representation of the fitness interactions within endosymbioses 16 1.2. Diagrammatic representation of the P. bursaria - Chlorella endosymbiosis 22 1.3. The consequence of symbiosis for each partner 25 2.1. Correlated metabolite enrichment for the 186b and HA1 P. bursaria and Chlorella strains over time 40 2.2. Fitness of the native and non-native host-symbiont pairings relative to isogenic symbiont-free hosts 44 2.3. Difference in Chlorella global metabolism between strains across light conditions 45 2.4. Difference in P. bursaria global metabolism between strains across light conditions 48 2.5. Photophysiology measurements for the native and non-native host-symbiont pairings 51 S2.1. PCR result of the HA1 and 186b Chlorella strains 56 S2.2. Schematic pathways diagram of nitrogen enrichment in the arginine amino acid metabolism of the Chlorella metabolic fraction 57 S2.3. Schematic pathways diagram of nitrogen enrichment in other aspects of amino acid metabolism in the Chlorella metabolic fraction 58 S2.4. Schematic pathways diagram of nitrogen enrichment in purine metabolism in the Chlorella metabolic fraction 59 S2.5. The interaction of light intensity and strain identity on the 13C enrichment profile of carbohydrate metabolites from the P. bursaria fraction. 60 3.1. Conceptual diagrams of potential outcomes when comparing native and non-native host-symbiont pairings 64 3.2. Initial growth rates of the host-symbiont pairings across a light gradient 69 3.3. Symbiont load of the host-symbiont pairings across a light gradient 70 3.4. The clustering of the metabolic fractions by light 72 3.5. Clustering patterns of the Chlorella metabolic fraction subset by host-genotype 73

3.6. Differences in the Chlorella metabolism between symbiont genotypes at multiple light levels within the 186b P. bursaria host 75 3.7. Relative abundances of dark-stress associated metabolites across host genotypes in the dark 80 3.8. Relative abundances of a high-light stress associated metabolite across host genotypes and across light levels 81 S3.1. PCR confirmation of symbiont-genotype within the reciprocal cross infections 87 S3.2. The clustering of the metabolic fractions by light in PCA plots 88 S3.3. Separation by symbiont-genotype within the 186b host subset of the Chlorella metabolic fraction 89 S3.4. Shared response of Chlorella genotypes to light intensity in the Chlorella metabolic fraction 90 4.1. Weekly growth rates of the native and novel symbioses across the evolution experiment 98 4.2. Growth rate assays performed at multiple points throughout the evolution experiment 99 4.3. Symbiont load at the start and end of the evolution experiment 100 4.4. Fitness of the host-symbiont pairings relative to the symbiont-free host at the start and end of the evolution experiment 101 4.5. The trajectories of the metabolic profiles from the start to the end of the evolution experiment. 103 4.6. Metabolites of interest across the start and end of the evolution experiment within the P. bursaria fraction. 104 4.7. Metabolites of interest across the start and end of the evolution experiment within the Chlorella fraction. 105

List of Tables

2.1 15N enriched metabolites of the Chlorella fraction 41 2.2 13C enriched metabolites of the P. bursaria fraction 42 2.3 The identified metabolites of interest from the Chlorella global metabolism. 46 2.4 The identified metabolites of interest from the P. bursaria global metabolism. 49 3.1 Symbiont-genotype specific metabolites in the dark within the 186b P. bursaria host 76 3.2 Symbiont-genotype specific metabolites in the intermediate light within the 186b P. bursaria host 77 3.3 Symbiont-genotype specific metabolites in the high light within the 186b P. bursaria host 78 S3.1 Light-level associated shared Chlorella metabolites across the host and symbiont genotypes. 91 S4.1 Identified metabolites associated with PCA trajectories for the P. bursaria fraction. 110 S4.2 Identified metabolites associated with PCA trajectories for the Chlorella fraction 112 S4.3 Change in symbiont load for each HK1 replicate between the start and end of the evolution experiment 113

Acknowledgements

First and foremost, I would like to thank my supervisors Michael Brockhurst, Duncan Cameron and A. Jamie Wood for making this project possible and extremely enjoyable. I have learnt a lot over the course of this PhD and it is due to their analytical guidance, which has taught me, among many things, an appreciation of the elegance of good research.

I would like to thank Ewan Minter for establishing many of the techniques used in this project and for taking the time to teach these to me. I also wish to thank Chris Lowe for his role in establishing this project and especially for his help while I conducted work in Falmouth.

I would like to thank Heather Walker for her technical expertise and help with the mass spectrometry.

I am grateful to the BBSRC White Rose DTP program for funding my PhD.

To the Brockhurst lab group, thank you for creating a culture that is scientifically exciting, supportive and fun. In particular, thank you to Ellie Harrison and Jamie Hall for your guidance and support. A special thank you to fellow officemates Cagla, Rosanna & Rachael whose friendship I value tremendously.

Lastly to my family, Mor, Far & Kim, you have been a constant source of support and inspiration, and thank you for always being willing to listen to how the algae was doing.

Declaration

I, the author, confirm that the Thesis is my own work. I am aware of the University’s Guidance on the Use of Unfair Means (www.sheffield.ac.uk/ssid/unfair-means). This work has not been previously been presented for an award at this, or any other university.

The following publications have arisen from this thesis:

• Sørensen, M.E.S., Lowe, C.D., Minter, E.J.A., Wood, A.J., Cameron, D.D., and Brockhurst, M.A. (2019). The role of exploitation in the establishment of mutualistic microbial symbioses. FEMS Microbiol Lett 366.

• Sørensen, M.E.S, Wood, A.J., Minter, E.J., Lowe, C.D., Cameron, D.D., and Brockhurst, M.A. (2020). Comparison of independent evolutionary origins reveals both convergence and divergence in the metabolic mechanisms of symbiosis. Current Biology.

Parts of chapter 1 are adapted from Sørensen et al. (2019), and the work in chapter 2 was published as Sørensen et al. (2020) following the submission of this thesis.

Chapter 1 Introduction

Parts of this chapter are adapted from a publication - The role of exploitation in the establishment of mutualistic microbial symbioses (Sørensen et al., 2019) (see Appendix A).

Endosymbiosis is the most intimate form of symbiosis, and therefore of interspecies interaction, as two unlike organisms live together with one organism residing within the cells of the other (De Bary, 1879). Endosymbioses can accelerate evolutionary innovation through the merger of once independent lineages, providing species with new ecological traits and allowing them to inhabit previously inaccessible ecological niches (Kiers and West, 2015; Wernegreen, 2012). The establishment of endosymbiosis can constitute a ‘major evolutionary transition’ (Szathmáry and Smith, 1995) in that previously autonomous entities merge, become mutually dependent, and form a new individual (Estrela et al., 2016; West et al., 2015). Endosymbiotic interactions include a vast array of diverse relationships across the three domains of life, and their formation has had extensive consequences for both the evolutionary history of life on Earth and its current ecological function. The primary endosymbiotic events that formed the mitochondria and plastids have shaped the evolution of eukaryotes and arguably enabled the emergence of complexity (Keeling, 2010; Martin et al., 2015). Ecologically, endosymbioses occur throughout the eukaryotic tree of life (Archibald, 2009) and by virtue of their adaptive evolutionary innovation, these associations often occupy keystone positions in ecosystems (Zook, 2002); for instance, plant-mycorrhizal associations form the main producers in most terrestrial ecosystems (Powell and Rillig, 2018), and coral-dinoflagellate associations form the foundation of coral reef ecosystems (Baker, 2003; Stanley and Lipps, 2011).

Transitions in individuality are, however, fraught with evolutionary conflict, and the merger of two independent organisms is rarely seamless and never selfless. Symbiosis encompasses a broad range of species interactions, including both parasitism (+/– fitness interactions) and mutualism (+/+ fitness interactions). Whilst the evolutionary rationale for parasitism is straightforwardly explained by the self-interest of the parasitic partner, explaining the origin of mutualistic symbiosis is more challenging (Frank, 1997; Sachs et al., 2004). The immediate fitness gains of cheating are expected to outweigh the potential long-term fitness benefits of cooperation, producing a ‘tragedy of the commons’ (Hardin, 1968; Rankin et al., 2007). Therefore, both in long-established associations and in the establishment of new relationships, evolutionary conflict and breakdown of mutualistic

11 symbiosis is ever likely, since each partner is under selection to minimise its investment in the integrated symbiotic unit (Perez and Weis, 2006; Sachs and Simms, 2006). Nevertheless, mutualistic symbiotic relationships are abundant, taxonomically widespread, ecologically important in a wide range of habitats, economically important in agricultural systems, and consequently underpin the biodiversity and function of both natural and man- made ecosystems (Bronstein, 2015; Powell and Rillig, 2018). It is the prevalence of mutualism in the face of evolutionarily conflict that fascinates researchers, and the mechanisms of mutualism maintenance continues to be a developing research area (Archetti et al., 2011; Douglas, 2008; Werner et al., 2018).

This introduction will briefly discuss the organelles as examples of the highest level of integration that endosymbionts have yet acquired, before focusing on secondary endosymbioses and their evolutionary dynamics. It then focuses on the endosymbiotic relationship between Paramecium bursaria and Chlorella, which provides a tractable experimental system for studying the evolution of endosymbiosis and is the focus of the experiments in this thesis. Finally, I outline the following data chapters and the questions they address.

1.1 The Organelles The organelles, the mitochondria and plastids, arose from primary endosymbiotic events that have subsequently shaped the course of life history (Keeling, 2010; Martin et al., 2015). The bacterial ancestry of the organelles was first proposed by Konstantin Mereschkowski (1905) and later championed by Lynn Margulis (1967), but remained controversial until molecular techniques became sufficiently advanced to provide unequivocal supporting evidence (Bonen and Doolittle, 1975; Schwarz and Kössel, 1980). Crucially, Bonen and Doolittle (1975) compared rRNA sequences between algal plastids and cyanobacteria to demonstrate the prokaryotic origin of these eukaryotic organelles, and shortly after, mitochondrial rRNA was also shown to be prokaryotic (Bonen et al., 1977). These results combined with an accumulated wealth of cytological, physiological and biochemical data (Dodson, 1979; Stanier, 1970) led to the acceptance of the serial endosymbiotic theory (Gray and Doolittle, 1982). Both mitochondria and chloroplast genomes are incredibly reduced, retaining a small fraction of their original complement of genes. They represent a very rare and highly derived subset within endosymbioses (Cavalier-Smith, 2013).

Mitochondria formed from α-proteobacteria endosymbionts within an archaeal host (Rivera and Lake, 2004; Spang et al., 2015), and this association lead to the formation of the

12 eukaryotic domain of life (Martin et al., 2015). Despite forming circa 1.5 BYA, mitochondria are remarkably persistent and though they have been reduced to hydrogenomsomes on multiple occasions (Allen et al., 2003), only in one instance have they ever been truly lost. This instance was facilitated by lateral gene transfer that enabled the nucleus to gain full metabolic independence (Karnkowska et al., 2016). Mitochondria are integrated with their hosts at an exquisite level of detail, to the extent that the complexes in their respiratory chain are a mosaic of proteins encoded in both the mitochondria and nucleus (Schatz and Mason, 1974).

Plastids, in comparison, are more transitory, and though stable organelles, their distribution throughout the eukaryotes is a complex mixture of acquisition, loss and replacement (Keeling, 2010). Plastids evolved from an endosymbiosis between a eukaryotic host and a cyanobacterium over a billion years ago that established photosynthesis in the eukaryotes (Dyall et al., 2004; Parfrey et al., 2011). Subsequently, multiple secondary endosymbioses, in which a eukaryotic host engulfs a plastid-bearing alga, spread plastids across the eukaryotic tree of life (Archibald, 2009; Keeling, 2013). The exact identity of these secondary endosymbioses is still being untangled, but what is certain is, unlike with mitochondria, plastid loss has happened on numerous occasions (Gornik et al., 2015). Furthermore, serial symbiont replacement appears to have occurred and the ‘shopping bag model’ hypothesises that the replaced symbiont can have transferred genes to the host leading to a complement of endosymbiont genes from mixed origins (Larkum et al., 2007; Patron et al., 2006). The complicated story of secondary gains and losses of plastids paints a far more fluid picture than that of mitochondria acquisition and is more representative of endosymbiosis as a whole; the relationship is maintained when necessary but lost when it is no longer advantageous.

Intriguingly, there has been a recent, 60-200 million years old (Berney and Pawlowski, 2006; Nowack et al., 2008), independent primary endosymbiosis — the eukaryote host Paulinella chromatophora acquired a cyanobacterial Synechococcus endosymbiont in a process that recapitulates the original evolution of the plastid (Marin et al., 2005; Nowack, 2014). The definition of an organelle currently requires protein import of a transferred gene back into the endosymbiont/organelle and this has not yet been conclusively demonstrated within the P. chromatophore endosymbiosis. Nonetheless, this relationship blurs the distinction between endosymbiont and organelle and has led many to question whether the current distinction between organelles and endosymbionts is meaningful. Or if, in fact, such

13 derived and integrated associations are already functionally equivalent to organelles (Bhattacharya and Archibald, 2006; Bodył et al., 2007; Keeling and Archibald, 2008).

1.2 Secondary Endosymbioses Organellegenesis is a special case of endosymbiosis, yet endosymbiosis more generally is a common evolutionary stable strategy with important evolutionary and ecological consequences. Endosymbiosis occurs between unrelated species, including between species that belong to different domains of life, and arguably the less related the partners are the greater the potential for acquisition of novel ecological traits (Douglas, 2014; Wernegreen, 2012). Eukaryote-bacteria endosymbioses are particularly common and span a wide range of functions and environments. For example: the chemoautotrophic endosymbionts of the giant worm Riftia pachyptila enable life at deep sea vents (Cavanaugh et al., 1981); the defensive Rickettsiella endosymbionts of pea aphids decrease predation (Tsuchida et al., 2010); the Nardonella endosymbionts cause cuticle hardening in weevils (Anbutsu et al., 2017); and Vibrio fischeri within Bobtail squid produce luminescence (McFall-Ngai and Ruby, 1998). Less common, and also less studied, are eukaryote-archaea endosymbioses; the majority of which have been observed within protist hosts, in particular many anaerobic ciliates possess methanogen archaea endosymbionts (van Hoek et al., 2000). Within-domain endosymbiosis can also introduce biological innovation. Eukaryote- eukaryote endosymbioses include the secondary acquisition of plastids that had spread photosynthesis across the eukaryotic taxons (Lane and Archibald, 2008) and the many endosymbioses between algae and animals, such as corals, sponges and cnidaria (Venn et al., 2008). Prokaryote-prokaryote endosymbioses are very rare, and to date have only been documented in nested endosymbiosis, such that the prokaryote host is itself an endosymbiont of an eukaryote. For instance, the mealybug Planococcus citri houses β- Proteobacteria that in turn houses γ-Proteobacteria (Dohlen et al., 2001). The scarcity of prokaryote-prokaryote endosymbioses, as opposed to their many ectosymbioses and syntrophic consortiums, is believed to be because of the absence of phagocytosis in prokaryotes (López-García et al., 2017).

Endosymbioses provide a multitude of functions, including the production of antibiotics (Currie et al., 1999), luminescence (Tebo et al., 1979), photoprotection (Hörtnagl and Sommaruga, 2007), and defence against predation and parasitism (Tsuchida et al., 2010). Nutritional endosymbioses are, however, the most common and show a higher degree of dependence when compared to defensive symbioses (Fisher et al., 2017). Nutritional endosymbioses occur across a broad range of taxa and can lead to highly integrated

14 metabolism, in which one biochemical pathway requires the complementation of enzymes from both partners. A classic example are aphids and their obligatory endosymbiont Buchnera aphidicola, which share the synthesis of the essential amino acids between them (Moran et al., 2003; Wilson et al., 2010).

Photosymbioses are associations where microalgae live within a heterotrophic host and therefore enable mixotrophy (Decelle, 2013). The transition to mixotrophy represents a fundamental shift in nutritional strategy that combines the roles of producer and consumer (Esteban et al., 2010; Stoecker et al., 2009), and consequently these associations often have important roles in ecosystems (Stanley and Lipps, 2011). These relationships are based on the transfer of fixed carbon from the photosynthetic partner in exchange for nitrogen and/or phosphate. Photosymbioses are extremely common and occur in a range of organisms, including a wide range of unicellular, protist hosts (Keeling, 2013; Stoecker et al., 2009) and multicellular organisms (Venn et al., 2008). Examples include: cyanobacteria and fungi in types of lichen (Honegger, 1991), haptophytes and Acantharia protist hosts (Decelle, 2013) and dinoflagellates and cnidarian in corals (Yellowlees et al., 2008).

1.3 The parasitism-mutualism continuum The fitness outcome of a symbiosis is determined by the balance of cost and benefit for each partner (Figure 1.1). The outcomes span from parasitism (+/– fitness interactions) to mutualism (+/+ fitness interactions) and form a continuum between these two states. A given symbiotic relationship is not a stationary point on this continuum because the benefits and costs of the symbiosis are dynamic and depend upon the environmental context, the stage of development and the interacting genotypes (Thompson, 2005). Indeed, because many of the potential benefits may only be required in particular environments or at particular times, the fitness outcome of many symbioses vary on ecological scales (Heath and Tiffin, 2007; Wendling et al., 2017). As such, some organisms only engage in symbiosis when in nutrient deficient environments (Johnson, 2011; Muscatine and Porter, 1977). The nature of symbiotic relationships is, therefore, context dependent. For example, increased nitrogen or phosphate fertilisation of the soil lowers the benefit of mycorrhizal fungal symbionts for a range of plant species, leading to reduced abundance of symbiotic interactions (Treseder, 2004). One consequence of context dependent fitness outcomes is that there is likely to be no universally optimal partner, which drives symbiotic relationships to be evolutionarily dynamic (Heath and Tiffin, 2007).

Figure 1.1. Diagrammatic representation of the fitness interactions within endosymbioses.

Mutualisms are defined by net positive fitness effects of interaction for both partners, but, as previously discussed, even mutualistic symbioses have an inherent potential for conflicting fitness interests among partners because of the short-term advantage of cheating. Mutualisms are, however, abundant throughout the tree of life despite their inherent evolutionary conflicts, and this disparity is considered the paradox of mutualism. Explaining the establishment of mutualistic symbioses is therefore challenging. The conditions for mutualistic symbioses to establish through mutualism alone are highly restrictive, and thus several alternative mechanisms have been proposed (Garcia and Gerardo, 2014; Keeling and McCutcheon, 2017). One of these is that mutualistic symbioses evolve from parasitisms. This transition can occur in two directions. First, the smaller parasitic partner living in or on the larger host can evolve reduced virulence to eventually become beneficial to its host (King et al., 2016; Shapiro and Turner, 2018; Tso et al., 2018). Sach et al. (2011) used phylogenetic reconstruction to predict whether bacterial symbionts originated as mutualists or parasites. For 42 beneficial bacterial symbionts, they inferred that 32 had originated as parasitic whilst only 9 had originated as mutualists (with 1 case remaining ambiguous), suggesting that parasitism is a more common route than mutualism to mutualistic symbiosis. Second, the larger host partner could capture and exploit the smaller beneficial partner, which would otherwise grow faster outside of symbiosis. This is a special case of parasitism known as host exploitation, which has been far less well-studied. This alternative route was proposed and modelled by Law & Dieckmann (1998), the model predicted that exploitative relationships can evolve into stable mutualistic symbioses with

16 vertical transmission simply through natural selection to increase individual fitness. The key requirement for this outcome was that the free-living symbiont pays a cost, which produces a trade-off for the symbiont. The symbiont either uses resources to overcome the cost of the free-living state or to provision the exploitative host but cannot do both. The model demonstrated that if the trade-off is sufficiently strong, the evolution of stable symbiosis can be advantageous to both partners even in an initially exploitative relationship.

To better understand the role for exploitation in the origin of mutualistic symbioses, there has been a recent call to reassess the fitness interactions of endosymbiotic relationships. Notably, Decelle (2013) has proposed that exploitation is likely to have been a common route for the origin of photosymbioses in particular. Currently there is evidence to suggest that the symbioses between scleractinian corals and the dinoflagellate algae Symbiodinium (Dubinsky and Berman-Frank, 2001; Smith and Muscatine, 1999; Wilkerson et al., 1988), the lichen symbiotic partners (Ahmadjian, 1993), chemosynthetic bacteria and their invertebrate hosts (Combes, 2005), and some protist-algal endosymbioses (Decelle, 2013; Lowe et al., 2016) are examples of host exploitation. Others go further, Keeling and McCutcheon (2017) state that endosymbioses are better viewed as “context dependent power struggles” or mutual exploitations, and that on evolutionary timescales conflict always remains.

1.4 Evolution of partner dependency In nature, the degree of dependence varies extensively both within and between symbioses (Fisher et al., 2017; Minter et al., 2018). Dependence ranges from obligate associations with mutually dependent partners, through asymmetrically dependent associations where only one species is unable to survive alone, to fully facultative associations where both species can survive alone. The potential asymmetry of dependence can cause conflict as one partner relies completely upon the other, while the other partner maintains the option of a free-living lifestyle.

Comparative studies suggest that mutual dependence is more likely to evolve in vertically- inherited symbioses, where the fitness interests of both species are more aligned compared to associations with some horizontal transmission. For reproductive interests to become fully aligned, both absolute co-dispersal and reproductive synchrony are required as part of vertical transmission (Frank, 1997). If achieved, this reduces within-host competition between symbionts and stabilises the symbiosis because the reproductive success of the

17 symbiont is perfectly correlated to that of its host. Vertical inheritance is common in well- established, obligate symbiotic partnerships and is associated with greater dependence (Fisher et al., 2017). It is not, however, ubiquitous and there are many stable mutualisms that maintain horizontal transmission. For example, Vibrio fischeri and bobtail squids (Visick and Ruby, 2006), Rhizobia and legumes (Sprent et al., 1987), and Endoriftia persephone and tube worms (Nussbaumer et al., 2006). Consequently, it is apparent that while vertical transmission helps to promote stability of some interactions, it is neither a necessary nor sufficient condition for the evolutionary stability of mutualistic symbioses (Genkai-Kato and Yamamura, 1999).

The evolution of mutual dependence is often associated with greater integration and genetic adaptation to symbiosis because these organisms no longer need to support a free-living life stage (Bennett and Moran, 2015). As such, once dependence has evolved integration of the partners extends beyond the initial function of the symbiosis. For example, after 150 million years of evolution aphids now rely on their obligate nutritional symbiont, Buchnera aphidicola, for a wide range of non-nutritional functions, including development even when dietary supplements are provided (Koga et al., 2007; Wilkinson and Ishikawa, 2000). Genome reduction is commonly seen in the genomes of anciently endosymbiotic taxa because, in the absence of a free-living life-stage, many genes are redundant; the symbiont resides in a stable host environment and can rely on the host to fulfil the majority of functions (McCutcheon and Moran, 2012). The extent of genome reduction can be extreme, and the smallest bacterial genomes are those of bacterial endosymbionts, for example the circada endosymbiont Candidatus Hodgkinia cicadicola has a genome of only 143,795 bps (McCutcheon et al., 2009). Host genomes will also alter, either in direct response to the symbiotic interaction, for instance the genes involved in provisioning the symbiont may be duplicated (Dahan et al., 2015), or because of endosymbiont gene transfer (EGT) the host genome may acquire new genetic material. To date there are only a few examples of EGT from non-organelle endosymbioses, including a number of Wolbachia-to-host-nucleus gene transfers with some of these transferred genes even being transcribed (Hotopp et al., 2007).

Although mutual dependency is associated with stable endosymbioses, in its most extreme form it can, however, cause an interaction to breakdown. For, once dependent, the host must maintain a relationship with a symbiont whose genome undergoes decay, becomes increasingly eccentric, and may lead the association to disappear down an evolutionary dead-end or “rabbit hole” (Bennett and Moran, 2015). In particular, the small effective population size and asexual nature of symbiont genomes mean they become increasing

18 subject to drift and so accumulate deleterious mutations. The extreme genome reduction of symbionts is likely to only be possible because the host functionally compensates for the decaying symbiont genome. The key symbiont genes, that encode the symbiotic functions upon which the host depends, normally remain under purifying and/or positive selection (Sabater-Muñoz et al., 2017). However, in the most extreme cases even these genes erode and then host compensation is not possible, in which case either the endosymbiosis goes extinct or symbiont replacement/supplementation must occur. Examples of the latter include the recurrent symbiont replacements of Hodgekina by entomopathogens in cicadas (Matsuura et al., 2018) and meadow spittlebugs whose ancient, highly reduced, symbiont has been replaced with a new Sodalis-like symbiont that has much higher genetic functionality (Koga and Moran, 2014). The new symbiont, however, faces the same evolutionary forces as the first and will likely also be subject to genome decay over evolutionary time.

1.5 Conflict avoidance A range of mechanisms have been proposed to ensure the maintenance of endosymbiosis in the face of evolutionary conflict and environmental variability. There are fundamental aspects of the relationship that can reduce conflict, known as ‘conflict avoidance factors’ (Herre et al., 1999). These include vertical transmission, genetic uniformity of symbionts, and obstructions to symbionts entering alternative free-living states. In addition, active mechanisms to police cheaters have been documented within mutualistic relationships and help to prevent the breakdown of these relationships. These include, partner sanctions in the legume-rhizobium symbiosis (Kiers et al., 2003), partner choice in the yucca-yucca moth symbiosis (Bull and Rice, 1991), partner fidelity in solitary wasp- Streptomyces symbiosis (Kaltenpoth et al., 2014), and screening in the bobtail squid-Vibrio fischeri symbiosis (Archetti et al., 2011; McFall-Ngai and Ruby, 1991).

Partner switching can terminate an association if its benefit-to-cost ratio is too low, enabling the acquisition of a new, more beneficial partner. This mechanism is predicted to be particularly effective in the context of environmental change or migration and niche expansion. Symbiont-mediated resilience to environmental change has been observed in corals that have acquired novel, thermally resistant Symbiodinium endosymbionts following thermal bleaching events (Boulotte et al., 2016); and niche expansion in lichens was enabled by replacement of photobiont ecotypes (Rolshausen et al., 2018). Partner switching is not always beneficial, however, and although theory predicts that low-benefit partners should be out-competed, a neutral partner can in theory become fixed within a

19 population (Fukatsu et al., 1994). The complexity of partner switching is shown by a recent phylogenetic analysis of the nutritional endosymbionts within hemipteran insects, which found that the replacement of the primary symbiont was related to adaptive dietary transitions, but that the more frequent turnover of secondary symbionts were not correlated to diet and may have been neutral (Bell-Roberts et al., 2019). Nonetheless, in some circumstances partner switching can provide rapid adaptation through the acquisition of novel traits (Gilbert et al., 2010).

1.6 Ecology and Physiology of the P. bursaria – Chlorella endosymbiosis Empirical data on the establishment of mutualistic symbioses are rare because studying this process experimentally is challenging. The extant mutualistic symbioses we observe in nature are the products of co-evolution and are no longer in the establishment phase. Furthermore, for obligate mutualistic symbioses it may be impossible to separate the partners and therefore untangle the costs/benefits that each of the symbiotic partners derive. Nevertheless, there are several beneficial microbial symbioses that are amenable to experimental study and are emerging as model systems for the study of symbiosis. Microbial systems are particularly powerful tools because their fast generation times, ease of laboratory culturing, high fecundity, and relatively smaller and easier to manipulate genomes make experimental procedures easier (Hoang et al., 2016; Jessup et al., 2004). In addition, microbial systems are a one-to-one symbiosis, and do not have the complications of working with a multicellular host.

One of the best studied facultative mutualistic endosymbioses is that between the ciliate Paramecium bursaria and the green algae Chlorella spp. This relationship has long been known to science, with P. bursaria formally being described in 1836 (Focke, 1836), and the vertical nature of the endosymbiosis described in 1960 (Siegel, 1960). While the physiology and ecology of this system have been studied for several decades, only recently are the molecular details of this interaction being uncovered. The relationship is experimentally tractable because it is easily culturable, with fast generation times and multiple laboratory techniques have been established for studying this system. Crucially, this association is usually facultative and therefore the consequences of symbiosis can be assessed separately for each partner. Consequently, this system has been used to experimentally test hypotheses on the evolution on endosymbiosis (Fujishima and Kodama, 2012).

Ciliates are a very diverse group of single-celled eukaryotes and are believed to have once carried plastids but lost them and reverted to heterotrophic lifestyles for the most part

(Reyes-Prieto et al., 2008). A multitude of diverse endosymbioses occur in ciliates (Fokin, 2004; Gast et al., 2009; Nowack and Melkonian, 2010), however, within Paramecium, only two species form photosymbiotic endosymbioses; whereas the association between P. bursaria and Chlorella is geographically widespread, the other between Paramecium chlorelligerum with Meyerella planctonica algae is much rarer (Kreutz et al., 2012; Lanzoni et al., 2016). P. bursaria cells are large, ~100µm across, and covered in cilia for motility and to draw food into the cells via the oral groove (Corliss, 1961; Fenchel, 1987). These host cells vastly dwarf their symbiont, despite both being unicellular eukaryotes, and one host will house between ~100-600 algal symbiont cells (Johnson, 2011; Kadono et al., 2004).

Chlorella are a large genus within the green algae, though the genus phylogeny and exact taxonomic group is still currently being defined, with ‘true’ Chlorella belonging to the Trebouxiophyceae class (Takeda, 1988). The cells are 2-10µm in diameter, house a single chloroplast, and reproduce both asexually and sexually (Blanc et al., 2010). Chlorella have a cellulose-glucosamine cell wall and in symbiotic Chlorella the cell wall is half the thickness of free-living cells (Higuchi et al., 2018). In recent years there has been a surge of interest in Chlorella and its potential applications, as a nutritional product (Rodriguez-Garcia and Guil-Guerrero, 2008), biofuel producer (Demirbas, 2011) and bioreactor (Walker et al., 2005), due to their high abundance and diversity of lipids and fatty acids (Converti et al., 2009; Safi et al., 2014). Intriguingly, Chlorella are very common symbionts and are found within amoeba, sponges, coelenterates (including Hydra), molluscs, flatworms and 25 species of ciliates (Zagata et al., 2016). It is unknown whether this alga has a propensity for symbiosis or if its multiple associations are simply a consequence of its abundance.

This endosymbiosis is primarily a nutritional symbiosis, centred upon the classical photosymbiotic exchange whereby fixed carbon from the photosynthetic Chlorella is exchanged for organic nitrogen from the heterotrophic P. bursaria (Figure 1.2) (Esteban et al., 2010; Reisser, 1976). In addition to this primary nutrient exchange, gas exchange also occurs as a beneficial by-product whereby the CO2 from P. bursaria respiration can be a substrate for photosynthesis in Chlorella, and the O2 from Chlorella photosynthesis can act as substrate for P. bursaria respiration (Johnson, 2011). Chlorella endosymbionts have been estimated to release 57% of their fixed carbon to the host cell (Johnson, 2011), primarily as maltose (Ziesenisz et al., 1981). In order to provide maltose both day and night two different pathways are utilised: in the light, maltose is synthesised de novo from the products of the Calvin Cycle, while, in the dark, maltose is formed from starch degradation (Ziesenisz et al., 1981).

Figure 1.2. Diagrammatic representation of the P. bursaria – Chlorella endosymbiosis.

Showing the nutrient exchange with the transfer of maltose from the Chlorella in exchange for organic nitrogen (denoted as ‘N’ as the identity of this compound is currently unknown). Ma = macronucleus; Mi = micronucleus.

The P. bursaria host acquires nitrogen from the digestion of bacteria (Johnson, 2011), and therefore must maintain its heterotrophic lifestyle even when housing autotrophic Chlorella. The identity of the nitrogen source provided by the host to algal symbionts is unknown, though multiple candidates have been proposed. The dominant theory is that nitrogen is provided as an amino acid; evidence supporting this comes from experiments demonstrating that the Japanese symbiotic Chlorella strain F36-ZK has lost its nitrate reductase activity but can utilise amino acids (Kato et al., 2006) and that symbiotic Chlorella strains grew better on urea or amino acids compared to inorganic nitrogen sources (Albers et al., 1982; Kessler and Huss, 1990). Furthermore, growth measurements of isolated symbiotic Chlorella on different nitrogen sources found that asparagine and serine supported growth in symbiotic but not free-living Chlorella, while other amino acids, including arginine and glutamine, could be utilised by both groups of Chlorella (Quispe et al., 2016). It is unclear, however, whether these patterns help to identify the exchange metabolite, because this compound need not necessarily be exclusively metabolised by symbiotic algae. In addition, results from Minaeva and Ermilova (2017) imply that arginine may be the transfer compound because the arginine concentration within symbiotic Chlorella matches that of isolated cells grown on arginine-supplemented medium, while much lower arginine concentrations occur in isolated cells grown on nitrate-based medium. Moreover, arginine supports growth of Chlorella as its sole N source (Arnow et al., 1953).

Alternatively, it has been proposed that P. bursaria’s nitrogen waste includes nucleic acid derivatives, such as guanine and xanthine (Soldo et al., 1978), and that these are then assimilated by Chlorella (Shah and Syrett, 1984). Nucleoside recycling occurs in other endosymbioses (Ramsey et al., 2010), and the utilisation of a host waste product would decrease the cost of symbiosis for P. bursaria. Additionally, there are conflicting results for ammonia utilisation, with some studies supporting it as a candidate nitrogen source (Albers et al., 1982) and others reporting poor Chlorella growth on ammonia-based media (Kato et al., 2006). The multiple, and somewhat conflicting, candidates for the nitrogen source could be explained if there is divergence among host-symbiont pairings, or if multiple nitrogen sources are provided simultaneously, alternatively further research may lead to a consensus around a single source.

The P. bursaria – Chlorella endosymbiosis utilises vertical inheritance of the symbiont, and synchronisation of their cell cycles (Kodama and Fujishima, 2012) and circadian clocks (Miwa et al., 1996). The division of the Chlorella is controlled by the host and occurs just prior to host cell division with a signal that is connected to the arrest of host cytoplasmic streaming (Takahashi et al., 2007). The circadian cycles of the symbiotic partners are interconnected, and the Chlorella sets the cycle for both partners (Miwa et al., 1996). Miwa et al. (1996) demonstrated that symbiotic P. bursaria have a longer clock period compared to aposymbiotic cells, that arrhythmic P. bursaria mutants can be rescued by symbionts, and that the host will shift in phase to match its Chlorella if out of sync. Furthermore, metabolic integration has occurred, and the nutrient exchange is actively regulated, for instance host Ca2+ inhibits serine uptake into Chlorella and glucose increases the uptake (Kato and Imamura, 2008a, 2008b). If the symbiont’s maltose is broken down to glucose by the host, then this control process would facilitate a reward system for co-operative symbionts.

Aposymbiotic P. bursaria are rarely isolated from natural populations (Tonooka and Watanabe, 2002), but experimental procedures for separating the partners have been developed. The host can be cured of symbionts through treatment with herbicide chemicals, such as paraquat and cycloheximide (Kodama and Fujishima, 2008) and the symbionts can be released from the host cells by sonication, which disrupts the host membrane (Kodama et al., 2014). There is strain variation in the level of dependency (Minter et al., 2018); among five geographically diverse isolates, both partners were fully facultative in some strains, while in others they displayed mutual obligacy or host obligacy. The P. bursaria were more dependent on the symbiosis than the Chlorella, such that only one

Chlorella strain of the five tested was incapable of free-living growth, while three of the P. bursaria strains were incapable of free-living growth. This suggests asymmetry in selection for dependency between the partners, consistent with the hypothesis that this association is based upon host exploitation of the algal symbiont and not mutual benefit (Lowe et al., 2016).

The facultative nature of the relationship has allowed the re-establishment of the symbiosis to be characterised in detail by Kodama & Fujishima (Kodama and Fujishima, 2011; Kodama et al., 2016). The Chlorella are engulfed along with food particles and initially contained in a digestive vacuole. Chlorella cells selected to form endosymbionts are partitioned and individually held in perialgal vacuoles that protect against lysosomal fusion. These are repositioned to just beneath the cell cortex to maximise light harvesting, in a similar fashion to chloroplast positioning within plant cells. However, the basis upon which Chlorella cells are selected to become endosymbionts is unknown. One theory is that the P. bursaria detect carbohydrate secretion by compatible symbiotic Chlorella. This is supported by the observation that Chlorella kept in the dark prior to inoculation will be digested rather than selected (Kodama and Fujishima, 2014) and that Chlorella maltose release is induced by low pH (Kamako and Imamura, 2006; Shibata et al., 2016), the environment of the perialgal vesicles within which Chlorella are held (Schüßler and Schnepf, 1992). Low-pH mediated carbohydrate secretion has also been observed in other phylogenetically distinct photosymbioses (e.g. those between Hydra and Chlorella (Douglas and Smith, 1984) and Dinoflagellates and coral (Tremblay et al., 2013)) suggesting perhaps that carbohydrate secretion is a commonly used cue for symbiosis-initiation. This apparently universal property may suggest that it is an ancestral physiological response of the algae that has been co-opted by the hosts as an ‘honest’ signal, rather than this being a symbiosis-specific adaptation.

The separation of the partners allows the fitness costs and benefits of symbiosis versus free-living to be directly quantified and compared. For hosts the benefit of symbiosis increased with light intensity, such that while it was costly to harbour symbiotic algae in the dark (i.e., symbiont-free hosts grow faster than symbiotic hosts), these costs were outweighed at higher light intensity such that symbiosis became highly beneficial relative to free-living for hosts in high light (Figure 1.3a) (Lowe et al., 2016). In contrast, symbiosis was never beneficial for the alga; free-living algal growth rates increased monotonically with light intensity and at all light levels exceeded those of symbiotic algae (Figure 1.3b). Furthermore, if the association is costly for an extended period of time the interaction can

24 breakdown, for instance complete darkness or chemical inhibitors of photosynthesis lead to the eventual loss of Chlorella symbionts through either digestion or egestion (Karakashian, 1963; Kodama and Fujishima, 2008). Endosymbioses are particularly susceptible to the shifts in the benefit-to-cost ratio during their establishment phase, and for the establishment to be successful it is likely that the light intensity would have to be above the no-benefit threshold regularly.

Figure 1.3. The consequence of symbiosis for each partner. A.) Host growth rate in response to light within symbiotic and cured P. bursaria (Figure 1A in the source). B.) Estimates for the photosynthetic efficiency (Fv/Fm) between symbiotic and isolated Chlorella (Figure 3A in the source). Responses are presented as the mean (n=3) ±SE. Adapted from Lowe et al., (2016) Current Biology.

Hosts manipulate the costs of symbiosis by regulating algal symbiont load (i.e. the number of algal symbionts per host cell), which consequently has a unimodal relationship with light intensity, peaking at low light, and being reduced both in the dark and at high light intensity (Lowe et al., 2016). A mathematical model of the symbiosis showed that hosts manipulate symbiont load in this way to maximise their return from nutrient trading, effectively minimising their nitrogen cost for each molecule of carbon they gain from their algal symbionts (Dean et al., 2016). Indeed, measurements of algal photosynthetic efficiency suggested that algal symbionts were more nitrogen starved than their free-living counterparts (Lowe et al., 2016). Similar patterns of benefit-to-cost ratio and host control were observed across a range of geographically diverse isolates (Minter et al., 2018). The packaging of Chlorella in host-derived vacuoles (Kodama and Fujishima, 2011, 2014) provides a clear mechanism of host control. Host regulation of symbiont load is believed

25 to arise through host-triggered symbiont division (Takahashi et al., 2007) and/or digestion/egestion of symbionts.

Taken together, the asymmetry in the benefit of the symbiosis and host-controlled regulation of symbiont load, suggest that the nutrient trading relationship between the ciliate and the alga is exploitative rather than mutualistic, benefiting the host (Lowe et al., 2016). Additional selective forces may be required therefore to explain the benefit, if any, of engaging in this symbiosis for the alga: both photoprotection and escape from viral predation have been proposed (Esteban et al., 2010; Reisser et al., 1991; Summerer et al., 2009). A cost in the free-living state, such as predation, could provide a sufficiently strong trade-off between the symbiotic and free-living state of the algae such that the evolution of stable symbiosis can be advantageous to both partners even in an exploitative relationship (Law and Dieckmann, 1998).

An important by-product of photosynthesis is photo-oxidative stress, predominantly in the form of damaging reactive oxygen species (ROS). Hundreds of photosynthesising Chlorella cells bring the potential for a vast increase in ROS, most of which will be contained within the Chlorella cells themselves, but hydrogen peroxide (H2O2) can cross membranes and may accumulate in the P. bursaria cytosol. Despite this potential, symbiotic P. bursaria have lower photo-oxidative stress and lower mortality rates than aposymbiotic cells at high UV (Hörtnagl and Sommaruga, 2007; Summerer et al., 2009). This suggests that not only do the Chlorella sufficiently protect the host from the ROS they produce, but that they provide additional protection for the host at high UV. The hosts nonetheless show behavioural responses to high light and will aggregate to create shading in high UV (Summerer et al., 2009). The relationship to stress within this relationship is complex, and Kawano and colleagues (Kawano et al., 2004) have hypothesised that P. bursaria were pre- adapted to photosymbiosis because they possessed a higher ROS tolerance than other Paramecium species, which allowed them to engage in this potentially lethal relationship. It is interesting to note, that ROS can play other biological roles besides causing damage:

H2O2 enables communication between chloroplasts and mitochondria, indicating that these compounds can be harnessed by the cell (Foyer and Noctor, 2003; Neill et al., 2002).

1.7 Genetics of the P. bursaria – Chlorella endosymbiosis Chlorella is a well-studied taxonomic group with a reasonably well-detailed genetic annotation and multiple genome sequenced species, including a P. busaria symbiotic type- strain, NC64A (Blanc et al., 2010). The NC64A genome reveals adaptations to symbiosis,

26 including increased numbers of genes involved in amino acid transport and carbon metabolism, compared to free-living Chlorella, traits that both relate to the core nutrient transfer of the symbiosis. Also enriched in the symbiont genome were protein families involved in protein-protein interactions which are hypothesised to be involved in symbiosis- specific signalling (Blanc et al., 2010). The NC64A genome also revealed orthologs of plant hormones, including abscisic acid, cytokinin and auxin receptors, which is in line with the increasing evidence that phytohormones are active in microalgae (Kiseleva et al., 2012; Tarakhovskaya et al., 2007).

P. bursaria has been less characterised at a genetic level, owing to the challenges of its genetic architecture. It possesses two nuclei: the micro nucleus, which is the inherited copy that acts as the germline, and the macro nucleus, a polyploid version of the genome that is actively transcribed (Corliss, 1961; Wichterman, 1986). An additional difficulty arises through the epigenetic modification between the two nuclei, involving the excision of almost all the transposable elements and internal eliminated sequences from the micro nucleus when the macro nucleus forms (Preer, 2000; Singh et al., 2014). Despite these complications, recent work has started to piece together the genetics of P. bursaria. The transcriptome of symbiotic versus aposymbiotic P. bursaria was compared by Kodoma et al. (2014). They found decreased carbon metabolism and host-mediated oxidative stress responses in symbiotic P. bursaria cells; both of these functions are expected to be partially taken over by symbiont metabolism. In addition, increased expression of histidine kinase and HSP70 in symbiotic P. bursaria cells, was suggested to be related to symbiosis coordination. Recently, an almost complete P. bursaria genome sequence was compared to a non-symbiotic close relative, Paramecium caudatum, by He et al. (2019). They found that P. bursaria encoded more genes related to nitrogen metabolism and that these genes were more highly expressed. In particular, the glutamine synthetase gene (glnA) had four times higher expression in P. bursaria than in P. caudatum, suggesting that glutamine may be the amino acid transferred to the algal symbiont. Alternatively, the increased expression may be reflective of increased nitrogen demand for downstream pathways that include the synthesis of other amino acids (glutamine synthetase being the primary route through which nitrate enters central metabolism (Rigano et al., 1981)). The P. bursaria genome also contained more genes encoding mineral absorption than the P. caudatum genome, and it has been hypothesised that Mg2+ levels could provide a mechanism for host-mediated symbiont load control, given that the chlorophyll compound is built around a Mg2+ ion. Furthermore, P. bursaria encoded fewer genes involved in oxygen binding than P.

27 caudatum, which may reflect redundancy given the ready supply of oxygen produced by Chlorella photosynthesis.

Analysis of symbiotic Chlorella nuclear rDNA loci, including 18S rDNA, ITS1 and ITS2, have revealed that symbiotic and free-living Chlorella form polyphyletic groups (Hoshina et al., 2005). The pattern of which strongly indicates that there have been multiple, independent origins of the P. bursaria symbiosis that involve different Chlorella species, although the exact number of symbiotic originations is currently unclear. However, a consistent pattern across multiple studies is that the P. bursaria-symbiotic Chlorella form two main biogeographical clades: a ‘European’ clade and a ‘American/Japanese’ clade (Hoshina and Imamura, 2008; Hoshina et al., 2005; Summerer et al., 2008). Within either of these two clades, the rDNA loci is highly conserved, but the rDNA sequences had characteristic intron insertions between the two groups, and the ITS2 sequences differed by almost 20% (Hoshina et al., 2004, 2005). The ‘European’ Chlorella clade associated with P. bursaria is more closely related to the symbiotic Chlorella of Hydra than it is to the ‘American/Japanese’ clade of P. bursaria-associated Chlorella, according to Hoshina et al., (2005). Despite this, host-species specificity has been demonstrated such that Chlorella from a non-ciliate host cannot successfully infect P. bursaria, including Chlorella from Hydra (Summerer et al., 2007). There is one example of an artificial initiation of a P. bursaria endosymbiosis with the cyanobacterium Synechocystis (Ohkawa et al., 2011), but there has been little follow up work on this intriguing interaction.

A phagotrophic protist such as P. bursaria feeds continually on bacteria, and, therefore, host-bacterial interactions happen continually. Most bacteria pass through the cell quickly, either being digested or escaping. Others are encased in vesicles for longer periods before digestion and are believed to be food storage vesicles. However, a few bacterial taxa appear to interact with the host and form stable endosymbioses. One potential example is Candidatus Sonnebornia yantaiensis, which lengthen P. bursaria survival if kept in pure water and locate close to the Chlorella perialgal vesicles (Gong et al., 2014). However, it is still debated whether they are true symbionts or simply long-term food stores (Gong et al., 2014). Across the Paramecium genus, almost 60 bacterial taxa have been reported as intracellular colonisers (Fokin, 2004). Though these additional relationships are yet to be thoroughly defined, P. bursaria seems likely to engage in other endosymbioses besides its core symbiotic relationship with Chlorella.

1.8 Thesis Outline

This thesis compares the multiple independent evolutionary origins of the P. bursaria - Chlorella endosymbiosis to understand the underpinning metabolic mechanisms. The chapters address the following specific questions:

Chapter 2: Comparison of independent evolutionary origins reveals both convergence and divergence in the metabolic mechanisms of symbiosis In this chapter I compared the metabolic mechanisms of two independent origins of the P. bursaria - Chlorella photosymbiosis using a novel reciprocal pulse-chase labelling experiment to reveal the pathways and dynamics of nutrient exchange. Predictions arising from the metabolic results were tested phenotypically with partner-switch experiments and physiological assays. These data suggest that the multiple origins of this symbiosis have a convergent mechanism of nutrient exchange, but that other important traits relevant to the host-symbiont phenotype have diverged between the independent origins of this endosymbiosis.

Chapter 3: Light-dependent stress-responses underlie host-symbiont genotypic specificity in a photosymbiosis Here I investigated the genetic variation for host-symbiont specificity in the P. bursaria - Chlorella endosymbiosis using a reciprocal cross-infection experiment coupled with metabolomics. The results reveal patterns of host-symbiont genetic specificity driven by contrasting light-dependent symbiont stress-responses.

Chapter 4: A novel host-symbiont interaction can rapidly evolve to become a beneficial symbiosis I experimentally evolved a novel host-symbiont pairing to test if initially non-beneficial associations formed through partner switching can evolve to become beneficial. Changes in host-symbiont growth rate, symbiont load, relative fitness, and metabolomics were quantified over time. The results show that the novel symbiosis could rapidly evolve to become equivalently beneficial to the native control through convergent metabolic mechanisms.

Chapter 5: Discussion I discuss the results of the three data chapters, synthesising the findings to provide an overall account of their implications for our understanding of the evolution of endosymbioses. In particular, I discuss the consequences of my results in the context of stress responses, partner switching and the rapid evolutionary adaptation of novel associations.

Chapter 2

Comparison of independent evolutionary origins reveals both convergence and divergence in the metabolic mechanisms of symbiosis

2.1 Introduction

Eukaryotic complexity is underpinned by endosymbiotic relationships, from the ancient mergers that led to the organelles, to the abundant and diverse secondary endosymbioses that provide novel metabolic capabilities across diverse taxa (Douglas, 2014; Moran, 2007). Many eukaryotes depend on their endosymbiotic partners for nutrition and survival (Ankrah and Douglas, 2018; Fisher et al., 2017; Johnson, 2011). The mechanisms that enable establishment of new associations have rarely been elucidated. This is partly because the origins of endosymbiotic relationships are difficult to study, but comparison of extant relationships can provide insight. In particular, where a symbiotic relationship has originated multiple times, these independent originations can be compared to determine the degree of convergence and divergence in the molecular mechanisms underpinning the symbiosis (Corsaro et al., 1999; Moran and Wernegreen, 2000; Sachs et al., 2011). Independent evolutionary origins of a beneficial symbiotic relationship suggest that a strong selective advantage has, on multiple occasions, overcome the inherent conflict between the self-interest of the partners. Independent origins of symbiosis appear to be common and have been reported for diverse symbiotic relationships, such as in lichens (Gargas et al., 1995; Muggia et al., 2011), aphids and their secondary symbionts (Sandström et al., 2001), the fungus–growing ant system (Munkacsi et al., 2004), and rhizobia-legume associations (Masson-Boivin et al., 2009).

The experimentally tractable microbial symbiosis between the ciliate host Paramecium bursaria and the algal endosymbiont Chlorella has arisen independently multiple times. This endosymbiosis relies on a classical photosymbiotic exchange between fixed carbon from the photosynthetic algae and organic nitrogen from the heterotrophic host (Johnson, 2011; Ziesenisz et al., 1981). This relationship has originated on at least two independent occasions giving rise to distinct geographical clades, known as the European clade and the

American/Japanese clade (Hoshina and Imamura, 2008; Summerer et al., 2008). This relationship has been well-characterised in regards to the establishment process and the integration of the partners (Fujishima, 2009; Kato et al., 2006; Kodama and Fujishima, 2011; Miwa et al., 1996). However, less is known about the symbiotic phenotypes and the mechanisms of convergence and divergence among the clades, except for their variation in nutritional requirements (Kamako et al., 2005; Kessler and Huss, 1990). Furthermore, it is unclear whether partner-switching can occur between the two main clades; with some studies indicating that it can (Summerer et al., 2007) but others suggesting that it cannot (Weis, 1978).

In photosymbioses the nutritional exchange provides the primary benefit of the symbiotic interaction, suggesting that this exchange is the crucial mechanism enabling the establishment of new associations and partner-switching (Decelle et al., 2015; Karkar et al., 2015). In the P. bursaria - Chlorella photosymbiosis the algae symbionts release 57% of their fixed carbon to their host, primarily as maltose (Ziesenisz et al., 1981). In exchange the host provides organic nitrogen, but the identity of the transferred nitrogen compound is unknown. Multiple candidates have been proposed and the dominant theory is that nitrogen is provided as an amino acid (Albers et al., 1982; Kato et al., 2006). The exact amino acid identity, however, has not be resolved because different studies have implicated different amino acids (He et al., 2019; Minaeva and Ermilova, 2017; Quispe et al., 2016). The nutritional exchange is a fundamental component of this endosymbiosis, and as such it has been found to be a critical aspect of the establishment process, with maltose secretion believed to be a cue for the initiation of this association (Douglas and Smith, 1984; Kodama and Fujishima, 2014; Tremblay et al., 2013).

Disentangling the contributions of each partner to the interlinked symbiotic metabolism is challenging. Isotopic enrichment is a valuable tool for discerning the origin of compounds transferred between the organisms. Using dual labelling the fate of multiple elements can be followed bidirectionally to track transfers between two partners; for instance, C13 and N15 isotopes can be used to track metabolic exchange in photosymbioses. Bulk isotopic enrichment has been used to detail the origin of metabolites in a number of symbioses, including sponges and their microbial communities (Achlatis et al., 2018; Shih et al., 2019), a novel algal-fungal endosymbiosis (Du et al., 2019), and myco-heterotrophic orchids and fungal symbionts (Cameron et al., 2006, 2008). An extension of this is the combination of enrichment analysis with mass-spectrometry that allows fine-scale pathway resolution of enrichment. This has been successfully used to study the C flux in the cnidarian-

32 dinoflagellate symbiosis (Matthews et al., 2018) and to study the C and N flux in the amino acids of a legume-rhizobium association (Molero et al., 2011).

Here, I extend the current metabolic methodologies applied to endosymbioses by using a reciprocal bidirectional pulse-chase experiment on global metabolism, which allowed the transferred C and N to be simultaneously tracked at an individual metabolite level. I employed an untargeted LC-ToF method to gain an overview of the metabolism rather than isolated pathways. Prior to metabolomic analysis, the symbiotic partners were separated, allowing the host and symbiont fractions to be analysed separately, which enables metabolism of each partner and the fate of exchanged metabolites to be determined. Using this approach, I compared the metabolic mechanisms of two independent origins of the P. bursaria - Chlorella photosymbiosis. Furthermore, I tested the implications of the metabolic results with partner-switch experiments and physiological assays to build an understanding of the causes and consequences of the metabolic mechanisms in these clades. The results revealed a convergent primary nutrient exchange, which enabled partner-switching. In contrast, divergence was observed in the metabolic mechanisms of light management, leading to differences in photophysiology between the strains and phenotypic mismatches in partner-switched associations. I discuss the consequences of these results for partner-switching and the evolution of endosymbioses.

2.2 Materials and Methods

Culturing conditions P. bursaria stock cultures were maintained at 25ᵒc under a 14:10 L:D cycle with 50 µE m-2 s-1 of light. The two natural strains used were: 186b (CCAP 1660/18) obtained from the Culture Collection for Algae and Protozoa (Oban, Scotland), and HA1 isolated in Japan and obtained from the Paramecium National Bio-Resource Project (Yamaguchi, Japan). The stocks were maintained by batch culture in bacterized Protozoan Pellet Media (PPM, Carolina Biological Supply), made to a concentration of 0.66 g L-1 with Volvic natural mineral water, and inoculated approximately 20 hours prior to use with Serratia marscesens from frozen glycerol stocks.

To isolate Chlorella from the symbiosis, symbiotic cultures were first washed and concentrated with a 11µm nylon mesh using sterile Volvic. The suspension was then ultra- sonicated using a Fisherbrand™ Q500 Sonicator (Fisher Scientific, NH, USA), at a power

33 setting of 20% for 10 seconds sonification to disrupt the host cells. The liquid was then spotted onto Bold Basal Media plates (BBM) (Stein, 1979), from which green colonies were streaked out and isolated over several weeks. Plate stocks were maintained by streaking out one colony to a fresh plate every 3/4 weeks.

Symbiont-free P. bursaria were made by treating symbiotic cultures with paraquat (10 µg mL-1) for 3 to 7 days in high light conditions (>50 µE m-2 s-1), until the host cells were visibly symbiont free. The cultures were then extensively washing with Volvic and closely monitored with microscopy to check that re-greening by Chlorella did not occur. Stock cultures of the symbiont-free cells were maintained by batch culture at 25ᵒc under a 14:10 L:D cycle with 3 µE m-2 s-1 of light and were given fresh PPM weekly.

Cross Infections Symbiont-free populations of the two P. bursaria strains were re-infected by adding a colony of Chlorella from the plate stocks derived from the appropriate strain. The re- greening process was followed by microscopy and took between 2-6 weeks. Over the process, cells were grown at the intermediate light level of 12 µE m-2 s-1 and were given bacterized PPM weekly.

Diagnostic PCR The correct algae genotype within the cross-infections was confirmed using diagnostic PCR. The Chlorella DNA was extracted by isolating the Chlorella and then using a standard 6% Chelex100 resin (Bio-Rad) extraction method. A nested PCR technique with overlapping, multiplex Chlorophyta specific primers were used as described by Hoshina et al. (2005). Standard PCR reactions were performed using Go Taq Green Master Mix (Promega) and 0.5µmol L-1 of the primer. The thermocycler programme was set to: 94ᵒc for 5min, 30 cycles of (94ᵒc for 30sec, 55ᵒc for 30sec, 72ᵒc for 60sec), and 5 min at 72ᵒc.

Fitness assay P. bursaria cultures, both the symbiotic cross-infections and symbiont-free cells, were washed with Volvic and resuspended in bacterized PPM. The cultures were then split and acclimated at their treatment light level (0,12,50 µE m-2 s-1) for five days. Cell densities were counted by fixing 360 µL of each cell culture, in triplicate, in 1% v/v glutaraldehyde in 96-well flat bottomed micro-well plates. Images were taken with a plate reader (Tecan Spark 10M) and cell counts were made using an automated image analysis macro in ImageJ v1.50i (Schneider et al., 2012). The competitions were started by setting up microcosms

34 that each contained 50:50 populations of green and white cells (with target values of 20 green cells and 20 white cells per ml) that were in direct competition. Cells were sampled on day 0 and day 7 on a flow cytometer and the proportion of green to white cells was measured and used to calculate the selection rate. Green versus white cells were distinguished using single cell fluorescence estimated using a CytoFLEX S flow cytometer (Beckman Coulter Inc., CA, USA) by measuring the intensity of chlorophyll fluorescence (excitation 488nm, emission 690/50nm) and gating cell size using forward side scatter; a method established by Kadono et al. (2004). The measurements were calibrated against 8- peak rainbow calibration particles (BioLegend), and then presented as relative fluorescence to reduce variation across sampling sessions. The re-establishment of endosymbiosis takes between 2-4 weeks, and this method was tested to ensure that the symbiont-free cells do not re-green over the course of the experiment.

Fluorimetry The cells were washed and concentrated with a 11µm nylon mesh using sterile Volvic and re-suspended in bacterized PPM. The cultures were then split and acclimated to their -2 -1 treatment light condition (12, 24 & 50 µE m s ) for five days. Fv/Fm, ΦPSII, and NSV values were measured by fast repetition rate fluorimetry (FastPro8, Chelsea instruments fluorometer (Oxborough et al., 2012) following the manufactures procedure. Cultures were dark acclimated for 15 minutes prior to measurements. For maximum quantum yield, measurements were repeated until Fv/Fm stabilized (typically 3-5minutes) and Fv/Fm then estimated as an average of 10 measurements. ΦPSII was measured in response to an actinic light source at sequentially increasing irradiances between 0 – 2908 PFD with 110 flashes of 1.1µs at 1µs intervals following standard green algae protocol. Peak emission wavelengths of the LED used for excitations was 450nm. Non-photochemical quenching was estimated by the normalised Stern-Volmer coefficient, defined as NSV = Fo’/Fv’ (McKew et al., 2013) and corrects for differences in Fv/Fm between samples.

Metabolomics Cultures were washed and concentrated with a 11µm nylon mesh using Volvic and re- suspended in bacterized PPM. The cultures were first grown for three days at 50 µE m-2 s-1 to increase cell densities, and then split and acclimated at their treatment light condition (6 & 50 µE m-2 s-1) for three days. For the sampling, the cultures were split into 3 treatment: the control, N15 enrichment by the addition of labelled Serratia marscesens 13 13 -1 (100µl per microcosm), or C enrichment by the addition of HC O3 (100 mg L ). The

35 cultures were sampled at four time points (0,2,6,8 hrs after the enrichment event). There were three biological replicates for each sampling event.

At each sampling event, the symbiotic partners were separated in order to a get P. bursaria and Chlorella metabolic fraction. The P. bursaria cells were concentrated with a 11µm nylon mesh using Volvic and then the P. bursaria cells were disrupted by sonication (20% power for 10 secs). 1ml of the lysate was pushed through a 1.6µm filter, which caught the intact Chlorella cells, and the run-through was collected and stored as the P. bursaria fraction. The 1.6µm filter was washed with 5ml cold deionized water, and then reversed so that the Chlorella cells were resuspended in 1ml of cold methanol, which was stored as the Chlorella fraction.

The samples were analysed with a Synapt G2-Si with Acquity UPLC, recording in positive mode over a large untargeted mass range (50 – 1000 Da). A 2.1x50mm Acuity UPLC BEH C18 column was used with acetonitrile as the solvent. The machine settings are listed in detail below:

Mass spectrometry settings: Polarity: positive Capillary voltage: 2.3 kV Sample Cone voltage: 20 V Source Temperature: 100ᵒc Desolvation temperature: 280ᵒc Gas Flow: 600 L hr-1 Injected volume: 5µl

Gradient information:

Time (mins) W ater (%) Acetonitrile (%)

0 95 5

3 65 35

6 0 100

7.5 0 100

7.6 95 5

The P. bursaria and Chlorella fraction were analysed separately. The xcms R package (Benton et al., 2010; Smith et al., 2006; Tautenhahn et al., 2008) was used for automatic peak detection by extracting the spectra from the CDF data files, using a step argument of 0.01 m/z. The automatically identified peaks were grouped across samples and were used to identify and correct correlated drifts in retention time from run to run. Pareto scaling was applied to the resulting intensity matrix.

Isotope analysis For the P. bursaria isotope analysis the 13C labelled samples were compared with the control, while for the Chlorella analysis the 15N labelled samples were compared to the control. In order to identify isotopic enrichment without user bias, I used Random Forest (RF) models to identify metabolites that associated with the isotope labelling. This is a machine-learning decision-tree based approach that produces powerful multivariate regression and is an established method for high-throughput biological data (Touw et al., 2013), including metabolomics (Hopkins et al., 2017). The isotope label was used as the response variable to regress against the metabolic profile of each sample. Each random forest model was run with 1000 iterations, and each RF analysis was run 500 times to account for uncertainty in the rank score. For each run, the rank score of the RF importance (measured as the mean decrease in Gini) was recorded for each m/z bin. The mean and standard error of the rank score was then calculated to assess the consistency of the variable importance. In total 4 RF models were analysed within each fraction, 1 per timepoint.

The rank score values were then compared between the strains. The high proportion of shared metabolites were selected and filtered to select those that had a higher relative abundance in the labelled fraction than in the control. From these, the profile of each candidate metabolite was manually checked for isotopic enrichment, and when a clear enrichment profile was present the monoisotopic mass was identified. The enrichment proportion of the isotopic masses to the monoisotopic mass was calculated, and the natural enrichment value within the control fraction was subtracted from the enrichment in the labelled fraction. Following this calculation, it was possible to determine if enrichment had occurred, and if so, the monoisotopic mass was considered a ‘mass of interest’.

Unlabelled analysis For the unlabelled, control fraction, the data was compared between the strains by calculating the log2(Fold Change) between the conditions (either between the strains

37 within each light level, or between the light levels within each strain) in a series of pairwise contrasts for each metabolite. Student T-tests were performed between the relative abundances of the paired comparisons. The Benjamini–Hochberg procedure was used to account for the high number of multiple P-value comparisons, with the false discovery rate set to 0.1 and 0.05 (Storey and Tibshirani, 2003) as highlighted in the volcano plots.

Identification of significant masses Masses of interest were investigated using the MarVis-Suite 2.0 software (http://marvis.gobics.de/) (Kaever et al., 2009), using retention time and mass to compare against KEGG (https://www.genome.jp/kegg/) (Kanehisa and Goto, 2000; Kanehisa et al., 2019) and MetaCyc (https://biocyc.org/) (Caspi et al., 2018) databases. The Metabolomics Standards Initiative requires two independent measures to confirm identity, which the combination of retention time and accurate mass achieves. This analysis therefore confirms level 1 identification.

Data Analysis Statistical analyses were performed in R v.3.5.0 (R Core Team, 2018) and all plots were produced using package ggplot2 (Wickham, 2016). Physiology tests were analysed by both ANOVA and ANCOVA, with light, host and symbiont identity as factors.

ΦPSII results were analysed with non-linear mixed effects models (nlme) with the nlme R package (Pinheiro et al., 2019). The ΦPSII data was fitted to an exponential decay function:

(bI) ΦPSII =ae

Where a is a normalisation constant and b is the rate constant. The nlme model included random effects by replicate on each parameter and fixed factors of host, symbiont and light factors that interacted with a following model reduction. Model fitting entailed starting with the most complex possible model, which was then compared to simpler models, and in the case that their explanatory power were equal, the most parsimonious model was chosen. See the supplementary statistics table for further details on the statistics used.

2.3 Results

The P. bursaria - Chlorella endosymbiosis has originated multiple times and forms two distinct biogeographical clades, specifically, the European clade and the American/Japanese clade (Hoshina and Imamura, 2008; Summerer et al., 2008). Using a representative of each – the strain 186b originally isolated in the UK and strain HA1 originally isolated in Japan (clade identity was confirmed by diagnostic PCR (Figure S1)) – I first tested whether these clades used convergent biochemical mechanisms of carbon (from the photosynthetic endosymbiotic Chlorella) for nitrogen (acquired by the protist host though the ingestion and digestion of free-living bacteria) exchange. To do this, I devised a novel, reciprocal, temporally-resolved, metabolomic pulse chase experiment. Using 15N-labelled bacterial necromass, I traced isotopic enrichment derived from N assimilated through P. bursaria digestion in Chlorella metabolites. In parallel, using 13C- lablled HCO3 I traced isotopic enrichment derived from C fixed by Chlorella photosynthesis in P. bursaria metabolites. This allowed the metabolic fate of resources exchanged between symbiotic partners to be quantified over time, allowing comparison of symbiotic metabolism between the strains.

Using Random Forest models to identify Chlorella metabolites that co-varied with 15N enrichment, I observed a shared isotopic enrichment response in 46% of metabolites (i.e. had a high-ranking score in both strains), suggesting that both Chlorella strains directed the exchanged nitrogen through central nitrogen metabolism in similar ways (Figure 2.1a). Similarly, I observed a shared 13C enrichment response in 75.12 % of P. bursaria metabolites, suggesting a high degree of convergence between the P. bursaria host strains in how they utilised the C derived from their algal symbionts (Figure 2.1b). Smaller proportions of metabolites showed an asymmetric response (i.e., were high-ranked in one strain but low-ranked in the other; for 15N enrichment, 20.55% in 186b Chlorella and 9.55% in HA1 Chlorella; for 13C enrichment 13.17% in 186b P. bursaria and 3.42% in HA1 P. bursaria), and there were subtle temporal differences in enrichment patterns between strains, suggesting only limited divergence in utilisation of exchanged metabolites has occurred between these host-symbiont clades.

Figure 2.1. Correlated metabolite enrichment for the 186b and HA1 P. bursaria and Chlorella strains over time. Each data point represents a metabolite. In each scatterplot the mean Random Forest rank order of each metabolite in the HA1 strain is plotted against the mean rank order of each metabolite in the 186b strain. The rank order value is positively correlated with magnitude of the enrichment signal. Timepoint is shown by the colour of each data point. A.) 15N enrichment in the Chlorella fraction. B.) 13C enrichment in the P. bursaria fraction. For both panels, the mean rank order is derived from multiple Random Forest analyses (n=500).

Co-enriched metabolites with the strongest enrichment over time were identified using LC- ToFMS (simultaneously resolving the monoisotopic mass and chromatographic retention time for each M/Z). For 15N co-enrichment in Chlorella (Table 2.1), I identified metabolites associated with the amino acid and purine pathways, which have both previously been suggested as probable N exchange metabolites in this symbiosis. Targeted pathway analysis indicated that an amino acid (probably arginine) is the more likely N exchange metabolite from P. bursaria to Chlorella in both clades (see supplementary results and Figure S2-S4). In addition, I observed co-enrichment in larger, N-rich metabolites, including chlorophyll precursors, which most likely represent the largest N-sinks for Chlorella, thus becoming 15 13 enriched in N as a function of N demand. For C enrichment in P. bursaria (Table 2.2), I identified metabolites involved in carbohydrate and lipid metabolism, suggesting that symbiont derived C was directed to carbon storage, as well as enrichment in central and amino acid metabolism, which are likely to have a high turnover of carbon and represent strong carbon sinks. For some carbohydrate storage metabolites, I observed stronger differences in 13C enrichment between light conditions in the 186b compared to the HA1 strain (Figure S5), indicating strain differences in the rate of flux through some of co- enriched pathways.

Table 2.1: 15N enriched metabolites of the Chlorella fraction. List of the identified metabolites found to be enriched with 15N in the Chlorella fraction in both the HA1 and 186b strain. This table includes their candidate identification, such as their detected mass and retention time as well as the main pathway the candidate compounds function within. ‘RF Time’ refers to the timepoint at which the metabolite was identified by the Random Forest Model.

RF Time Detected Mass Retention Time Pathway Candidate Compounds Exact Mass Adduct KEGG/ MetaCyc 1 113 482 Pyrimidine/Amino acid Uracil 112.0273 H+ C00106 1,3-diaminopropane 74.0844 K+ C00986 1 166 478 Purine 5-Amino-4-imidazole carboxylate 127.0382 K+ C05516 1,2 237.1 286 Biotin Dethiobiotin 214.1317 Na+ C01909 1,2,3,4 871.6 405 Chlorophyll Pheophytin A 870.5659 H+ C05797 1,2,4 593.3 405 Chlorophyll Pheophorbide A 592.2686 H+ C18021 Urobilinogen 592.3261 H+ C05790 2,3 140 213 Amino acid L-Aspartate 4-semialdehyde 117.0426 Na+ C00441 Indole 117.0578 Na+ C00463 1-Aminocyclopropane-carboxylate 101.0477 K+ C01234 5-Aminopentanal 101.0841 K+ C12455 3 482.4 324 Folate biosynthesis Dihydrofolate 443.1553 K+ C00415 3 848.6 294 Ubiquinone Rhodoquinone-10 847.6842 H+ CPD-9613 4 227.1 460 Amino acid/Chlorophyll Tryptophan 204.0899 Na+ C00078 Porphobilinogen 226.0954 H+ C00931

Table 2.2: 13C enriched metabolites of the P. bursaria fraction. List of the identified metabolites found to be enriched with 13C in the P. bursaria fraction in both the HA1 and 186b strains. This table includes their candidate identification, such as their detected mass and retention time as well as the main pathway the candidate compounds function within. ‘RF Time’ refers to the timepoint at which the metabolite was identified by Random Forest Model.

RF Time Detected Mass Retention Time Pathway Candidate Compounds Exact Mass Adduct KEGG 1 100 16 Glycerophospholipid Ethanolamine 61.0528 K+ C00189 1 689.2 16 Carbohydrate Glycogen 666.2219 Na+ C00182 1,2 124 15 Vitamins and Cofactors Niacin 123.032 H+ C00253 1,2 261 14 Carbohydrate Monosaccharide phosphate 260.0297 H+ C00092 1,2,3 251 17 Isoprenoid pathway (R)-5-Phosphomevalonate 228.0399 Na+ C01107 1,2,3,4 190 341 Phosphonate Demethylphosphinothricin 167.0347 Na+ C17962 1,2,3,4 441.3 310 Lipid Hydroxycholesterol 402.3498 K+ C05500 1,2,3,4 639.2 414 Heme biosynthesis Haem 616.1773 Na+ C00032 1,2,3,4 212.9 479 Chlorocyclohexane and Chlorodienelactone 173.972 Ka+ C04706 chlorobenzene degradation 1,2,4 109 479 Quinone p-Benzoquinone 108.0211 H+ C00472 1,2,4 345.9 480 Amino acid metab 3-Iodo-L-tyrosine 306.9705 K+ C02515 1,3,4 169 19 Central metabolism 2-Oxoglutarate 146.0215 Na+ C00026 2-Oxoisocaproate 130.063 K+ C00233 3-Methyl-2-oxopentanoate 130.063 K+ C00671 2-Dehydropantoate 146.0579 K+ C00966 3-Phosphonopyruvate 167.9824 H+ C02798 Phosphoenolpyruvate 167.9824 H+ C00074 2 313.2 287 Lipid HPODE 312.2301 H+ C04717 2,3,4 519.1 400 Peptide Nitro-hydroxy-glutathionyl- 496.1264 Na+ C14803 dihydronaphthalene 2,4 71.1 373 Amino acid Aminopropiononitrile 70.0531 H+ C05670 3 405.1 236 Isoprenoid pathway Farnesyl diphosphate 382.131 Na+ C00448

The pulse-chase analysis suggests that these P. bursaria - Chlorella strains, representing independent origins of the symbiosis, show convergent utilisation of partner-derived nutrients, and I hypothesised therefore that partner-switched host-symbiont pairings would be functional. To test this, I performed a reciprocal cross-infection experiment whereby the P. bursaria host strains were cured of their native algal symbiont, and subsequently re- infected with either their native algal symbiont or the reciprocal non-native algal symbiont. I then directly competed each host-symbiont pairing against its respective symbiont-free host strain across a light gradient. I used flow cytometry to quantify the proportion of green (with symbiont) versus white (symbiont-free) host cells at the start and end of the growth cycle to calculate the selection rate, thus providing a direct measure of the fitness effect of symbiosis for hosts. As predicted, all the symbiont pairings showed a classic photosymbiotic reaction norm, such that the relative fitness of hosts with symbionts versus hosts without symbionts increased with increasing irradiance (Figure 2.2), and more steeply in the HA1 host background (host genotype * light environment interaction, ANOVA, F3,31 = 29.34, P< 0.001). This confirms that both host genotypes could derive the benefits of symbiosis from either of the symbiont genotypes, but that the fitness effect of symbiosis varied between strains.

Figure 2.2. Fitness of the native and non-native host-symbiont pairings relative to isogenic symbiont-free hosts. Lines show mean (n=3) competitive fitness of symbiont- containing hosts relative to their isogenic symbiont-free host genotype calculated as selection rate, and the shaded area denotes ±SE. The left-hand panel shows data for the HA1 P. bursaria, the right-hand panel the data for the 186b P. bursaria containing either native (solid line) or non-native (dotted line) Chlorella symbionts. Colour denotes the Chlorella genotype. Selection rate = 0 represents equal fitness. 43

These light-dependent differences in the fitness of the host-symbiont pairings suggest that the HA1 and 186b strains may have diverged in aspects of their metabolism and physiology outside of the primary symbiotic nutrient exchange. Next, to characterise potential differences in global metabolism between the HA1 and 186b host-symbiont strains, I performed untargeted metabolomics analyses on the unlabelled metabolites from the separated Chlorella and P. bursaria fractions. Pair-wise contrasts, both between the strains and between the light levels, were used to identify masses of interest (Figure 2.3 & 2.4). I observed a range of metabolites that differentiated the 186b and HA1 Chlorella strains (Table 2.3), and metabolism differed more between strains than it did between light conditions within strains (Figure 2.3). Notably, the HA1 Chlorella strain displayed higher levels of several carotenoids than the 186b Chlorella strain, particularly at high irradiance, whereas the 186b Chlorella strain displayed higher levels of metabolites involved in chlorophyll and ubiquinol metabolism than the HA1 Chlorella strain at both low and high irradiance. Fewer metabolites distinguished the global metabolism of the P. bursaria strains (Table 2.4). In all cases these metabolites were present at higher levels in the 186b P. bursaria strain compared to the HA1 P. bursaria strain (Figure 2.4), and neither strain’s metabolism varied significantly with irradiance (Figure 2.4). The identified metabolites that distinguished the strains were associated with a range of functions, including amino acid metabolism, amino sugars, and sphingolipid metabolism. Several other metabolites, although present in the host fraction, are likely to have been secreted into the host cytoplasm by the algal symbiont or be derived from the bacterial necromass. These include a zeatin candidate, which may play a role in Chlorella signalling, and several metabolites identified as putative antibiotics.

Figure 2.3. Difference in Chlorella global metabolism between strains across light conditions. Represented as volcano plots with the fold change of each metabolite against its statistical significance. The data points are highlighted at two false discovery rate (FDR) values, and if the Log2(fold change) is greater than 1 or less than -1. A.) Comparing the expression between the two strains within the high light condition. B.) Comparing the expression between the two strains within the low light condition. C.) Comparing expression between the two light levels within the HA1 strain. D.) Comparing expression between the two light levels within the 186b strain.

Table 2.3: The identified metabolites of interest from the Chlorella global metabolism. These metabolites were highlighted by the volcano plot (Figure 2.3) and had significantly higher abundances in either one of the strains or one of the light conditions within the Chlorella fraction. Detected Retention Kegg / Upregulated in Condition Mass Time FDR Pathway Candidate Compounds Exact Mass Adduct Metacyc HA1 strain H & L light 247.2 336 *,** Alkaloid/quinone Anapheline 224.1889 Na+ C06183 Geranylhydroquinone 246.162 H+ C10793 283.3 336 *,** Fatty acid Oleate 282.2559 H+ C00712 H light 218.2 17 * Amino acid L-Glutamylputrescine 217.1426 H+ C15699 Alanyl-L-lysine 217.1426 H+ C05341 265.3 337 * Fatty acid 1-Hexadecanol 242.261 Na+ C00823 385.2 375 * Plant Hormone Gibberellin A36 362.1729 Na+ C11862 571.5 435 * Carotenoid Methoxyneurosporene 570.4801 H+ C15895 589.4 420 * Carotenoid Echinenone 550.4175 K+ C08592 Anhydrorhodovibrin 566.4488 Na+ C15877 Hydroxychlorobactene 550.4175 K+ C15911 3-Hydroxyechinenone 566.4124 Na+ C15966 591.4 420 * Carotenoid Zeaxanthin 568.428 Na+ C06098 Zeinoxanthin 552.4331 K+ C08590 beta-Cryptoxanthin 552.4331 K+ C08591 Xanthophyll 568.428 Na+ C08601 Low Light HA1 strain 743.5 373 * Phosphoglyceride 1-18:3-2-trans-16:1-phosphatidylglycerol 742.4785 H+ CPD-2186 186 Strain H & L light 105 15 *,** Central metabolism Hydroxypyruvate 104.011 H+ C00168 Allophanate 104.0222 H+ C01010 169 17 ** Central metabolism 2-Oxoglutarate 146.0215 Na+ C00026 Phosphoenolpyruvate 167.9824 H+ C00074 3-Phosphonopyruvate 167.9824 H+ C02798 2-Oxoisocaproate 130.063 K+ C00233 3-Methyl-2-oxopentanate 130.063 K+ C00671 2-Dehydropantoate 146.0579 Na+ C00966 Coumarin 146.0368 Na+ C05851 273.2 395 ** Fatty Acid 16-Hydroxypalmitate 272.2351 H+ C18218 289.3 244 ** Diterpenoid Kaurenol 288.2453 H+ C11872

Table 2.3 continued Detected Retention Kegg / Upregulated in Condition Mass Time FDR Pathway Candidate Compounds Exact Mass Adduct Metacyc 186 Strain H & L light 337.3 380 ** Fatty acids 13;16-Docosadienoic acid 336.3028 H+ C16533 607.3 361 ** Chlorophyll Protoporphyrinogen IX 568.305 K+ C01079

781.6 471 ** Ubiquinone 3-methoxy-4-hydroxy-5-nonaprenylbenzoate 780.2 H+ CPD-9898 925.6 359 ** Chlorophyll Bacterio-pheophytins 888.5765 K+ C05798 H light 262.1 248 ** Folate Dihydrobiopterin 239.1018 Na+ C00268 6-Lactoyl-5;6;7;8-tetrahydropterin 239.1018 Na+ C04244 323.2 248 * Photoreception Vitamin A aldehyde 284.214 K+ C00376 335.3 372 ** Isoprenoids Phytol 296.3079 K+ C01389 2-Octaprenyl-3-methyl-5-hydroxy-6-methoxy- 751.5 366 ** Ubiquinone 712.5431 K+ C05815 1;4-benzoquinone L light 273.3 268 ** Diterpenoid Ent-Kaurene 272.2504 H+ C06090

Figure 2.4. Difference in P. bursaria global metabolism between strains across light conditions. Represented as volcano plots with the fold change of each metabolite against its statistical significance. The data points are highlighted at two false discovery rate (FDR) values, and if the Log2(fold change) is greater than 1 or less than -1. A.) Comparing the expression between the two strains within the high light condition. B.) Comparing the expression between the two strains within the low light condition. C.) Comparing expression between the two light levels within the HA1 strain. D.) Comparing expression between the two light levels within the 186b strain.

Table 2.4: The identified metabolites of interest from the P. bursaria global metabolism. These metabolites were highlighted by the volcano plot (Figure 2.4) and had significantly higher abundances in either one of the strains or one of the light conditions within the P. bursaria fraction.

Upregulated Detected Retention Exact in Condition Mass time FDR Pathway Candidate Compounds mass Adduct KEGG 186 strain H & L light 124 238 **, * Vitamins and Cofactors Niacin 123.032 H+ C00253 126 217 **, * Sulfur metabolism Taurine 125.0147 H+ C00245 170 237 **, * Amino acid Glutamate 147.0532 Na+ C00025 5-Amino-4-oxopentanoate 131.0582 K+ C00430 Glutamate 5-semialdehyde 131.0582 K+ C01165 364.2 236 *, * Antibiotic ? ACV 363.1464 H+ C05556 396.1 237 *,* Antibiotic ? Deacetylcephalosporin C 373.0944 Na+ C03112 Novobiocic acid 395.1369 H+ C12474 H light 352.2 237 * Plant hormone? trans-Zeatin riboside 351.1543 H+ C16431 390.1 237 * Amino and nucleotide sugar N-Acetylneuraminate 9-phosphate 389.0723 H+ C06241 416.1 250 ** Antibiotic ? Cephalosporin C 415.1049 H+ C00916 Chlorobiocic acid 415.0823 H+ C12471 434.1 249 * Antibiotic ? Novobiocic acid 395.1369 K+ C12474 L light 418.2 268 * Sphingolipid metabolism Sphingosine 1-phosphate 379.2488 K+ C06124

The clear differences in global metabolism between the algal strains suggests that they may vary in their photophysiology, which could, in turn, help to explain the light- dependent differences in fitness observed in the reciprocal cross-infection experiment. To test this, I measured several key photochemical parameters in the native and non-native host-symbiont pairings. For two measures of photosynthetic efficiency — Fv/Fm (the intrinsic efficiency of photosystem II [PSII], Figure 2.5a) and ΦPSII (the proportion of the light absorbed by chlorophyll associated with PSII that is used in photochemistry, Figure 2.5b) (Maxwell and Johnson, 2000) — I observed a significant host genotype by symbiont genotype by light environment interaction (for FvFm ANOVA, F7,232 = 86.41, P<0.001; for

ΦPSII nlme model intercept summary ANOVA, F11,24 = 11.66, P<0.001 (see Appendix B for full statistical output)). In the HA1 P. bursaria host, the pattern of photosynthetic efficiency across the light gradient did not vary with algal strain, whereas in the 186b P. bursaria host, the native 186b Chlorella showed lower photosynthetic efficiency than the HA1 Chlorella at low growth irradiance, but the pattern was reversed at high growth irradiance. Correspondingly, the HA1 Chlorella produced more carotenoids at high irradiance than the 186b Chlorella, and carotenoids perform a role in photoprotection and can therefore decrease the light energy that reaches the photosystems and thereby limit photosynthesis.

Non-photochemical quenching is used by photosynthetic organisms to safely deal with excess and potentially damaging light energy and was estimated using the normalised Stern-Volmer coefficient (NSV). The NSV response (Figure 2.5c) across the actinic light gradient was significantly affected by host genotype for the intercept value suggesting differences among the host genotypes in their ability to photo-protect algal symbionts

(ANOVA, F1,34 = 4.74, P<0.05). Meanwhile, both symbiont genotype and growth irradiance affected the first coefficient (ANOVA, F3,32 = 5.56, P<0.01); and symbiont genotype affected the second coefficient (ANOVA, F1,34 = 8.932, P<0.01) (see Table S2.1 for full statistical output). Higher levels of NSV and steeper NSV reaction norms for the 186b Chlorella, particularly in its native host background, are consistent with the greater investment in photosynthetic machinery observed in the metabolome, allowing this genotype to better dissipate excess light energy as heat whilst not compromising photosynthetic efficiency.

Figure 2.5. Photophysiology measurements for the native and non-native host- symbiont pairings. For all subplots, lines represent the mean (n=3) and the shaded area denotes ±SE. In each subplot the left-hand panel shows data for the HA1 P. bursaria host, the right-hand panel shows data for the 186b P. bursaria host containing either native (solid) or non- native (dashed line). Colour denotes Chlorella genotype (186b in blue; HA1 in green). A) Estimates of the maximum quantum yield of photosystem II (Fv/Fm) across growth irradiances. B) Light-adapted quantum yield of photosystem II (ΦPSII) across growth irradiances, lines represent exponential decay models using nlme package in R. C.) The normalised Stern- Volmer quenching coefficient (NSV = Fo’/Fv’) across growth irradiances, presented at polynomial models. See Appendix B for model details.

2.4 Discussion

In this chapter, I have compared the metabolic mechanisms underpinning two independent origins of the P. bursaria - Chlorella photosymbiosis using a novel reciprocal metabolomic pulse-chase method. This showed highly conserved patterns of nutrient exchange and utilisation for both the host-derived N in the Chlorella genotypes and the symbiont-derived C in the P. bursaria genotypes. Consistent with a conserved primary symbiotic nutrient exchange, partner-switched host-symbiont pairings were functional. By directly competing symbiotic hosts against isogenic symbiont-free hosts, I showed that the fitness benefits of symbiosis to hosts increased with light irradiance but varied according to host genotype. 51

Global metabolism varied more strongly between the Chlorella than the P. bursaria genotypes and suggested divergent mechanisms of light management. Specifically, the algal symbiont genotypes either produced photo-protective carotenoid pigments at high irradiance or more chlorophyll and ubiquinol, resulting in corresponding differences in photosynthetic efficiency and non-photochemical quenching among host-symbiont pairings. These data suggest that the multiple origins of the P. bursaria - Chlorella symbiosis relied upon a conserved mechanism of nutrient exchange, whereas other traits linked to photosynthesis and thus the functioning of the photosymbiosis are divergent.

Reciprocal nutrient exchange is central to the P. bursaria - Chlorella symbiosis, however, whilst the carbon exchange metabolite has long been identified as maltose (Ziesenisz et al., 1981), the identity of the nitrogen transfer compound has thus far been unknown. Previous work has reported evidence supporting a role for amino acids (Kato et al., 2006; Kessler and Huss, 1990), but with conflicting information regarding the amino acid responsible and whether there are multiple transfer compounds. For example, a recent genomic comparison found that P. bursaria had 4 times higher expression of a glutamine synthetase gene (GlnA) than the non-symbiotic Paramecium caudatum (He et al., 2019). In contrast, Quispe et al. (2016) found that symbiotic but not free-living Chlorella could utilise asparagine and serine, whereas other amino acids, including arginine and glutamine, could be utilised by both symbiotic and free-living Chlorella. Although it is unclear that the exchange metabolite need necessarily be one exclusively metabolised by symbiotic algae. In contrast, the metabolomics analysis presented here indicated arginine as the most likely exchange metabolite in both genotypes based on the enrichment pattern, although our first sampled time-point was too late to detect the actual exchange metabolite. Nevertheless, this matches results from Minaeva and Ermilova (2017) who found the arginine concentration within symbiotic Chlorella matches that of isolated cells grown on arginine-supplemented medium, while much lower arginine concentrations occur in isolated cells grown on nitrate- based medium. Moreover, arginine supports growth of Chlorella as its sole N source (Arnow et al., 1953). The challenge in identifying the nitrogen exchange is that the metabolism of amino acid compounds is closely connected, especially for glutamine and arginine, which makes separating the true transfer compound from up/down-stream effects difficult. Future isotope enrichment experiments will be required to measure the enrichment profile more intensively over a shorter time-period.

This chapter has demonstrated highly conserved patterns of nutrient exchange and utilisation among two independent origins of the P. bursaria - Chlorella symbiosis, and

52 further showed that this enables partner-switching between clades. In contrast, Paramecium has previously been shown to be unable to establish symbiosis with algae isolated from other host species, such as Hydra (Summerer et al., 2007). This suggests that other photosymbioses may use alternative nutrient exchanges, which in turn prevent between-host partner-switches. This could be tested by comparing the nutrient exchange metabolism in other photosymbioses using the methods established here. Partner switching within a host species leads to the potential of symbiont replacement if multiple symbiont strains are locally available. Hoshina et al. (2012) demonstrated that co-infection of multiple algal symbionts within the same P. bursaria host cell is possible, using Chlorella variabilis and Micractinium reisseri (Chlorellaceae). Partner-switching can rescue symbioses from cheater-induced extinction by restoring symbiotic function (Koga and Moran, 2014; Matsuura et al., 2018), enable rapid adaptation to environmental change (Boulotte et al., 2016; Lefèvre et al., 2004), and facilitate niche-expansion (Joy, 2013; Rolshausen et al., 2018; Sudakaran et al., 2017). Local adaptation by symbiont acquisition is likely to occur far faster than by symbiont evolution and may be a general mechanism of ecological innovation in symbioses. For instance, it is believed to have enabled diversification in insect endosymbioses (Sudakaran et al., 2017). Furthermore, symbiont replacement is thought to have been an important factor in plastid evolution (Keeling, 2010) and ‘the shopping bag model’ hypothesises that serial symbiont replacement not only altered symbiont identity, but led to a complement of endosymbiont genes and proteins from multiple origins (Larkum et al., 2007). This arises because the preceding transient symbionts can have transferred genes to the host nucleus, which entangles the lineages. (Dorrell and Smith, 2011; Patron et al., 2006; Stiller et al., 2014).

The biogeographical clades have been defined at a molecular level (Hoshina et al., 2005) but there is very little work on their phenotypic differences beyond their nutritional requirements. Previous work identified strain variation in vitamin B12 requirement and inorganic nitrogen utilisation of symbiotic Chlorella (Kamako et al., 2005; Kessler and Huss, 1990), the latter may be due to divergent genome reduction following specialisation on host-derived organic nitrogen sources. This chapter identified significant differences between the strains across a range of relevant phenotypes, including, their global metabolism and photosynthetic responses. Our data suggest metabolic mechanisms for the observed differences in photosynthetic responses. The 186b Chlorella invests more than the HA1 Chlorella in the components of its photosystems, through chlorophyll and ubiquinol, providing 186b Chlorella with sufficient electron transport machinery to deal with excess light energy. This enables effective non-photochemical quenching, and high NSV values,

53 without compromising photosynthetic efficiency and results in high fitness of the 186b host-symbiont pairing at high light. This high photosynthetic investment strategy is likely to be costly, however, potentially explaining the higher cost of 186b Chlorella symbionts when in the dark. On the other hand, the HA1 Chlorella display higher levels of carotenoid pigments at high irradiance, which are likely to facilitate photoprotection and nonphotochemical quenching. However, this occurs at the expense of photosynthetic efficiency because these pigments decrease the amount of light energy that reaches the photosystems, and results in lower fitness of the HA1 symbiont-186b host pairing at high light. This mechanism is only expressed at high irradiance, suggesting that HA1 Chlorella adopts a responsive protection strategy to deal with high light intensity. Divergence in the metabolism of light management appears to provide the mechanistic basis for the variation in phenotype among strains and, therefore, may explain strain variation.

Direct measurement of the fitness effect of symbiosis is highly challenging in most associations, and consequently fitness is usually implied indirectly from growth rates or other traits believed to correlate with fitness (Heath and Tiffin, 2007). Here, I used a novel relative fitness assay that directly competes symbiotic hosts against isogenic symbiont-free hosts across a light gradient over several generations. This enables direct estimation of selection coefficients, and therefore of the fitness effects of symbiosis. The HA1 host gained a greater fitness benefit from symbiosis than the 186b host, regardless of symbiont genotype. The HA1 symbiosis is more likely, therefore, to be able to re-establish symbiosis. The 186b symbiosis is particularly costly in dark and low light conditions, and would therefore be likely to only re-establish symbiosis under high light conditions where symbiosis is beneficial. The differences between the strains, therefore, extends to their evolutionary fitness that in turn will determine when these endosymbioses can establish and under which conditions they are maintained.

Partner switching requires compatibility between host and symbiont if it is to rescue the breakdown of symbiosis (Boulotte et al., 2016; Matsuura et al., 2018), but to enable adaptation to new niches (Joy, 2013; Rolshausen et al., 2018) it requires phenotypic variation among symbiont genotypes. In this chapter I have shown that both of these characteristics exist within independent originations of the P. bursaria - Chlorella symbiosis. The results revealed the metabolic and phenotypic consequences of independent originations of symbiosis and showed that despite these differences, partner switching is possible because of evolutionary convergence to a shared nutrient-exchange. The concurrent divergence in the algae strain photophysiology altered the light-dependent

54 responses of the symbiosis, and similar genotype-dependent light responses have been observed in other photosymbioses (Abrego et al., 2008; Howells et al., 2012; Ye et al., 2019), suggesting that this may be an important cause of genotype by genotype interactions within photosymbiotic associations. The influence of partner identity on the symbiotic phenotype indicates that symbiont switching could potentially enable adaptation. Multiple independent originations occur in a diverse range of symbioses (Masson-Boivin et al., 2009; Muggia et al., 2011; Sandström et al., 2001) and this may be a critical input of genetic variation that enables adaptation to changing environmental conditions.

2.5 Supplementary Figures

186b HA1 L

1000 bp

500 bp

Figure S2.1. PCR result of the HA1 and 186b Chlorella strains. Overlapping, multiplex primers were used to amplify fragments within the 18S rDNA and ITS region of the Chlorella nuclear genome. In this region the ‘American/Japanese’ strains, such as HA1, have had three introns inserted that the ‘European’ strains, such as 186b, lack (Hoshina and Imamura, 2008; Hoshina et al., 2005). The banding patterns here match the expected pattern in that the HA1 fragment is considerably larger than the main fragment of 186b, and both have additional smaller fragments. Shown alongside a 100bp ladder.

Figure S2.2. Schematic pathways diagram of nitrogen enrichment in the arginine 15 amino acid metabolism of the Chlorella metabolic fraction. The tables show relative N enrichment across time (hrs) in the two light conditions (H = 50 µE m-2 s-1, L = 6 µE m-2 s-1). The colour corresponds to the fold change of the enrichment compared to the control, with significance stars indicating the statistical strength of this change. These results are further discussed in the Supplementary Results section.

Figure S2.3. Schematic pathways diagram of nitrogen enrichment in other aspects of amino acid metabolism in the Chlorella metabolic fraction. This data shows the amino acid metabolism that includes lysine, aspartate and serine. The tables show relative 15N enrichment across time (hrs) in the two light conditions (H = 50 µE m-2 s-1, L = 6 µE m-2 s-1). The colour corresponds to the fold change of the enrichment compared to the control, with significance stars indicating the statistical strength of this change. These results are further discussed in the Supplementary Results section.

Figure S2.4. Schematic pathways diagram of nitrogen enrichment in purine metabolism in the Chlorella metabolic fraction. The tables show relative 15N enrichment across time (hrs) in the two light conditions (H = 50 µE m-2 s-1, L = 6 µE m-2 s-1). The colour corresponds to the fold change of the enrichment compared to the control, with significance stars indicating the statistical strength of this change. These results are further discussed in the Supplementary Results section.

Figure S2.5. The interaction of light intensity and strain identity on the 13C enrichment profile of carbohydrate metabolites from the P. bursaria fraction. For all panels, the enrichment value is the Log2 of the Fold Change in enrichment of the 13C labelled fraction compared to the control, presented as the mean (n=3) ±SE. The low light level refers to 6 µE m-2 s-1 and the high light to 50 µE m-2 s-1. A) Profile of 689.2 mz, 16 rt, Glycogen. B) Profile of 365.1 mz, 16 rt, a disaccharide, thought to be sucrose.

2.6 Supplementary Results

Metabolic pathway analysis

Given that the low molecular weight compounds in the results of the 15N co-enrichment in Chlorella (Table S2.1) were almost exclusively amino acid or purine related, I focused on these pathways for a further targeted approach. Key compounds of these pathways were selected and searched for in the metabolite dataset. Overall, 16 potential components of amino acid metabolism were identified and 10 potential components of purine metabolism. To follow the flow of enriched nitrogen in these pathways, the enrichment profile of these compounds was calculated, and the results plotted as heatmaps, based on the method used by Austen et al. (In Press).

Within amino acid metabolism the nitrogen enrichment is focused downstream from arginine (Figure S2.2); ornithine, putrescine and citrulline possessed clear enrichment profiles while upstream compounds such as arginosuccinate had no detectable enrichment. Furthermore, other aspects of amino acid metabolism, such as that centred around aspartate, serine or lysine (Figure S2.3), showed little and inconsistent enrichment. Unfortunately, I could not identify a candidate compound for arginine to test if it had the enrichment profile of a transfer molecule (predicted to be a very high initial enrichment that then substantially decreased over time). Such a pattern was not seen for any compound, I suggest, therefore, that our first timepoint was not early enough to capture the initial enrichment events involving the transfer compound itself.

Within purine metabolism, the nitrogen enrichment occurred both up and downstream of the purine bases (Figure S3.4). The enrichment upstream of the purine bases indicates that enriched nitrogen is entering this pathway from the amino acid of central metabolism. Based on this pattern, I believe that the purine pathway is a site of secondary enrichment and it reveals that purine-derivatives present a substantial nitrogen demand.

Chapter 3

Light-dependent stress-responses underlie host-symbiont genotypic specificity in a photosymbiosis

3.1 Introduction

Photosymbioses are mixotrophic interactions whereby a heterotrophic host is provided with carbon fixed by intracellular photosynthetic microalgae (Decelle, 2013; Esteban et al., 2010). This important energetic transition often leads to photosymbioses performing unique, keystone functional roles in ecosystems. For example, the association between Symbiodinium and cnidarian hosts form photosynthetic coral that are the foundation of reef ecosystems (Baker, 2003). Photosymbioses are widely distributed throughout the tree of life, and occur both within microbial hosts (Keeling, 2013; Lane and Archibald, 2008) and metazoans (Venn et al., 2008). The nutritional exchange provides the primary benefit of photosymbioses and the photosynthetic partner normally receives nitrogen and potentially other key nutrients in return for the fixed carbon they provide. As a result of this exchange photosymbioses are often assumed to be mutualistic, but detailed experimental studies have reported instances of host exploitation (Decelle, 2013; Lowe et al., 2016). Photosymbioses are hugely important, but there are many aspects of these relationships we do not fully understand. In particular, we urgently need to understand how genetic variation affects the outcome of these host-symbiont interactions, since this is the fuel for their coevolution (Heath and Stinchcombe, 2014).

Genetic variation for the outcome of symbiosis, either in symbiotic traits or fitness, can be quantified as the host genotype by symbiont genotype interaction (GH × GS), also termed intergenomic epistasis (Heath, 2010). GH × GS interactions have been reported for a wide taxonomic range of symbioses. For example, the symbiont density of Wolbachia in its bean beetle host is affected by both host genotype and Wolbachia genotype (Kondo et al., 2005); transmission success of an oomycete pathogen in Arabidopsis thaliana depends on the specific combinations of host and parasite strain (Salvaudon et al., 2005); and aphid

62 performance on Trifolium is dependent on the genotype of its nutritional endosymbiont Regiella insecticola and aphid genotype (Ferrari et al., 2007).

It is well-established that the dynamics and outcomes of coevolution depend on the environmental context (the Geographic Mosaic theory) (Thompson, 2005). This is because the outcome of host-symbiont interaction is frequently environmentally context dependent, causing a host genotype by symbiont genotype by environment (GH × GS × E) interaction. While GH × GS × E interactions are common in host-parasite relationships (e.g., Wendling et al., 2017; Zouache et al., 2014), they are of particular importance in beneficial symbioses because variation along environmental axes related to the symbiotic exchange can shift the nature of the interaction along the parasitism-mutualism continuum. In both plant- mycorrhizae (Pieulell et al., 2008) and plant-rhizobia (Heath et al. 2010; Heath &Tiffin 2007) interactions, extensive variation exists in the host-symbiont response to environmental conditions. For example, Heath et al. (2010) reported variation in nodulation between N-fixing rhizobia and legume strains in response to environmental nitrate, such that increasing nitrate led some genotype combinations to reduce nodulation, others to increase nodulation, while some were unaffected. This case illustrates that the environment is fundamental to the function of symbioses, and, in certain combinations of environment and genotypes, the rhizobia – legume symbiosis was shifted to such an extent that it was more advantageous for the plant to partially dissociate than to continue the relationship. One consequence of context dependence is that there is likely to be no universally optimal partner, and consequently symbiotic relationships are evolutionarily dynamic (Heath and Tiffin, 2007).

Stress responses play an important role in the fitness of photosymbioses due to their exposure to potentially-damaging light energy (Venn et al., 2008; Yakovleva et al., 2009). Photo-oxidative stress is a by-product of photosynthesis that can cause damage to cells through reactive oxygen species (ROS) if it is not mitigated (Murata et al., 2007). Stress tolerance is likely to show genetic variation among host-symbiont associations. In coral - Symbiodinium endosymbioses, the symbiont genotype primarily determines the thermal and light tolerance of the association (Abrego et al., 2008; Howells et al., 2012), although the host genotype does also influence this process (Baird et al., 2009; Loya et al., 2001). If the stress tolerance of the symbiont is exceeded, then the symbiosis breakdowns and coral bleaching occurs (Weis, 2008). It has, however, been theorised that coral bleaching may potentially be adaptive as it provides an opportunity for the host to acquire a new symbiont genotype that is better adapted and more tolerant of the prevailing environmental

63 conditions (Buddemeier and Fautin, 1993; Gilbert et al., 2010). In the Hydra - Chlorella photosymbiosis, although algal genotype had some effect, the threshold of thermal tolerance was determined by the host genotype (Ye et al., 2019). In the microbial photosymbiosis between the heterotrophic ciliate Paramecium bursaria and the green algae Chlorella sp, symbiotic hosts have been shown to have lower photo-oxidative stress and lower mortality rates than aposymbiotic cells at high UV (Hörtnagl and Sommaruga, 2007; Summerer et al., 2009). Stress tolerance can therefore be affected by both partners of the photosymbiosis to varying degrees, but host-symbiont pairings that lack sufficient combined stress tolerance are unlikely to survive and will be prone to breakdown.

Despite the high level of genetic variation within the P. bursaria - Chlorella association, owing to the multiple independent originations of the symbiosis (Hoshina and Imamura, 2008; Summerer et al., 2008), there have been no systematic studies of the genotype by genotype by environment interaction within this endosymbiosis. Furthermore, the photo- oxidative stress response has not been studied in detail and has not been compared across strains nor across light gradients. Metabolomics can be used to identify the metabolic markers of stress while also examining central metabolism, and therefore this technique provides a tractable experimental system with which to study genetic variation within a photosymbiosis.

Figure 3.1. Conceptual diagrams of potential outcomes when comparing native and non-native host-symbiont pairings. The colour of the line denotes whether the symbiont is the native or non-native symbiont. A) Shows the ‘no effect’ outcome when there is no significant difference between native and non-native pairs. B) Shows the ‘native advantage’ outcome whereby the native pair has the advantage in all conditions. C) Shows the ‘interaction’ outcome where the native pair is advantageous is some conditions, but disadvantageous in others.

In this chapter, I investigated the genetic variation in the P. bursaria - Chlorella endosymbiosis using a reciprocal cross-infection experiment coupled with metabolomics. Potential outcomes of the cross-infections are visualised in Figure 3.1 and show how comparisons between the native and non-native host-symbiont pairs can reveal the degree of partner specificity. If, for instance, the host is a generalist then there should be no effect of symbiont identity (Figure 3.1a), while if the host is a specialist it may be that the native symbiont is always the most advantageous (Figure 3.1b) or that the benefit-to-cost ratio of the different symbionts is dependent on environmental conditions (e.g. a GH × GS × E interaction)(Figure 3.1c). I assessed the outcome of the cross-infections using phenotypic assays of host-symbiont growth rate and symbiont load, and investigated the global differences in metabolism using ESI-ToF untargeted metabolomic analysis. The results revealed a GH × GS × E interaction for the host-symbiont growth rate and the regulation of symbiont load. Moreover, I observed metabolic differences between the symbionts that offer potential mechanistic bases for host-symbiont specificity. Chiefly, that contrasting stress responses between the symbiont genotypes played an important role and may have altered the benefit-to-cost ratio of symbiosis for the host. I discuss how the differences in stress management may influence host-symbiont specificity and the implications for partner switching.

3.2 Materials & Methods

Cultures & Strains P. bursaria – Chlorella cultures were maintained under the conditions described in Chapter 2. The three natural strains used in this chapter were: 186b (CCAP 1660/18) obtained from the Culture Collection for Algae and Protozoa (Oban, Scotland), and HA1 and HK1 isolated in Japan and obtained from the Paramecium National Bio-Resource Project (Yamaguchi, Japan).

Cross infection The separation of symbiotic partners was achieved by the method described in Chapter 2. Once separated, the three aposymbiotic P. bursaria strains were re-infected by adding a colony of Chlorella from the plate stocks derived from the appropriate strain. This was done with all three of the isolated Chlorella strains to construct all possible host-symbiont genotype pairings (n=9). The re-establishment of endosymbiosis was confirmed on the

65 microscope and took between 2-6 weeks. Over the process, cells were grown at the intermediate light level of 12 µE m-2 s-1 and were given bacterized PPM weekly.

Growth rate Growth rates of the symbioses were measured across a light gradient. The cells were washed and concentrated with a 11µm nylon mesh using sterile Volvic and re-suspended in bacterized PPM. The cultures were then split and acclimated to their treatment light condition (0, 12, 24, & 50 µE m-2 s-1) for five days. The cultures were then re-suspended in bacterized PPM to a target cell density of 150 cell mL-1. Cell densities were measured at 0, 24, 48 and 72 hours by fixing 360 µL of each cell culture, in triplicate, in 1% v/v glutaraldehyde in 96-well flat bottomed micro-well plates. Images were taken with a plate reader (Tecan Spark 10M) and cell counts were made using an automated image analysis macro in ImageJ v1.50i (Schneider et al., 2012).

Symbiont load The symbiont load was measured in cultures derived from the growth rate experiment so that the data could be integrated between the two measurements. Triplicate 300µl samples of each cell culture were taken from 72 hour cultures for flow cytometry analysis. Host symbiont load was estimated using a CytoFLEX S flow cytometer (Beckman Coulter Inc., CA, USA) by measuring the intensity of chlorophyll fluorescence for single P. bursaria cells (excitation 488nm, emission 690/50nm) and gating cell size using forward side scatter; a method established by Kadono et al. (2004). The measurements were calibrated against 8-peak rainbow calibration particles (BioLegend), and then presented as relative fluorescence to reduce variation across sampling sessions.

Metabolomics Cultures of the symbiotic pairings were washed and concentrated with a 11µm nylon mesh using sterile Volvic and re-suspended in bacterized PPM. The cultures were then split and acclimated at their treatment light condition (0, 12 & 50 µE m-2 s-1) for seven days. The metabolomic fractions of the P. bursaria and Chlorella were separated with the method described in Chapter 2. After which the Chlorella fraction samples were already in methanol, but the P. bursaria fraction samples had then to be diluted by 50% with methanol. Metabolic profiles were recorded using ESI ToF-MS, on the Qstar Elite with automatic injection using Waters Alliance 2695 HPLC (no column used), in positive mode. This is an established high-throughput method with a large mass range (50 Da to 1000 Da).

Mass spectrometry settings: Polarity: positive Ion Spray voltage: 4.2 kV Declustering potential: 120 V Focusing potential: 265 V Source temperature: 200ᵒc Gas Flow: 40 ml min-1 Solvent: 50:50 methanol to water at flow rate 40µl min-1 Injected volume: 10µl

The processing was performed using in-house software Visual Basic macro 216 (Overy et al., 2005), which combined the spectra across the technical replicates by binning the crude m/z values into 0.2-unit bins. The relative mass abundances (% total ion count) for each bin was summed. Pareto scaling was applied to the results, and the data was then analysed by principal component analysis using SIMCA-P software (Umetrics). When treatment- based separation was observed, supervised orthogonal partial least squares discriminant analysis (OPLS-DA) separation was then performed using the discriminatory treatment with the SIMCA-P software.

For the pair-wise comparison of the metabolic profiles of interest, the log2(Fold Change) of relative abundance between the compared samples was calculated per metabolite. T- tests were performed on the relative abundance of the two samples, and the P-values were plotted against the log2(Fold Change) in the volcano plots.

Identification of significant masses Masses of interest were annotated using the initial identifications from the in-house software program and further comparisons against KEGG (https://www.genome.jp/kegg/) (Kanehisa and Goto, 2000; Kanehisa et al., 2019) and Metlin (https://metlin.scripps.edu) (Smith et al., 2005) databases. The Metabolomics Standards Initiative requires two independent measures to confirm identity, this analysis only used one measure (accurate mass) and therefore, meets only the level 2 requirements of putative annotated compounds.

Data Analysis Statistical analyses were performed in R v.3.6.1 (R Core Team, 2018) and all plots were produced using package ggplot2 (Wickham, 2016). Details of the statistical methods used are within the supplementary statistics table.

3.3 Results

Using three strains of the P. bursaria - Chlorella endosymbiosis representing both the ‘European’ (the 186b strain) and ‘American/Japanese’ (the HA1 and HK1 strains) clades, I constructed all possible host-symbiont genotype pairings (n = 9) and confirmed correct algal identity by diagnostic PCR (Figure S3.1). To determine host-genotype versus symbiont-genotype contributions to host-symbiont growth performance I measured the growth reaction norm of each host-symbiont pairing across a light gradient (Figure 3.2). All host-symbiont pairings showed the classic photosymbiotic reaction norm, such that growth rate increased with irradiance, but I observed a significant host-genotype by symbiont-genotype by light environment (GH × GS × E) interaction for host-symbiont growth rate (ANOVA, F17,162 = 18.81, P<0.001). This was driven by contrasting effects of symbiont genotype on growth in the different host backgrounds across light environments. In the HK1 and HA1 host-backgrounds similar growth reaction norms with light were observed for each symbiont genotype, whereas in the 186b host background the growth reaction norm varied according to symbiont genotype. Interestingly, the native 186b host- symbiont pairing had both the lowest intercept and the highest slope, indicating that in the 186b host background the native symbionts were costlier in the dark yet more beneficial in high-light environments than non-native symbiont-genotypes.

Figure 3.2. Initial growth rates of the host-symbiont pairings across a light gradient.

The data points show the mean (n=3) initial growth rate ±SE. Each panel shows the data for a specific genotype of P. bursaria host and host genotype is also represented by the shape of the data points. The symbiont genotypes are distinguished by colour.

P. bursaria host cells regulate their native symbiont load according to light irradiance to maximise the benefit-to-cost ratio of symbiosis, such that symbiont load peaks at low irradiance and is reduced both in the dark and at high irradiance (Dean et al., 2016; Lowe et al., 2016). To test if regulation of symbiont load varied among host-symbiont pairings, I measured symbiont load across a light gradient as the intensity of single-cell fluorescence by flow cytometry (Figure 3.3). All host-symbiont pairings showed the expected unimodal symbiont load curve with light, but I observed a significant GH × GS × E interaction for symbiont load (ANOVA, F17,162 = 3.78, P<0.001). Polynomial models were used to distinguish how the symbiont load curves varied among the host-symbiont pairings across the light gradient (plotted in Figure 3.3). The model coefficients showed a significant GH S x G interaction (ANOVA, F8,36 =27.22 (the intercept); 8.58 (first coefficient); 6.09 (second coefficient), P<0.001 (see Appendix C for full statistical output)). Whereas, in the HA1 host similar symbiont load reaction norms were observed for each symbiont genotype, for the HK1 and 186b host backgrounds the form of the symbiont load reaction norms varied according to symbiont genotype. In the HK1 host the magnitude of the symbiont load (Y at maximum) varied by symbiont genotype, such that higher symbiont loads were observed

for the native compared to the non-native symbiont-genotypes. In the 186b host, peak symbiont load (X at maximum) occurred at different light levels according to symbiont genotype, such that for the native symbiont the symbiont load curve peaked at a higher light intensity when compared to the non-native symbionts. This suggests that the HK1 and 186b host-genotypes discriminate among symbiont-genotypes, and then regulate symbiont load accordingly.

Figure 3.3. Symbiont load of the host-symbiont pairings across a light gradient. The data points show the mean (n=3) symbiont load, measured as relative chlorophyll fluorescence, ±SE. The lines show the polynomial models the data was modelled by; for full model details see

Appendix C. Each panel shows the data for a specific genotype of P. bursaria host and host genotype is also represented by the shape of the data points. The symbiont genotypes are distinguished by colour.

To investigate the potential metabolic mechanisms underlying the observed GH × GS × E interactions for host-symbiont growth rate and symbiont load, I performed untargeted global metabolomics with ESI-ToF-MS independently for the host and symbiont metabolite fractions for each host-symbiont pairing across the light gradient. Across the entire dataset, for both the host and symbiont metabolite fractions, principal component analysis revealed an overall pattern of clustering by light level (OPLS-DA Figure 3.4, PCA Figures S3.2). This suggests that light irradiance was the primary driver of differential metabolism for both host and symbiont, with broadly similar metabolic responses to light observed across all host-symbiont pairings. This shared response to light intensity was

70 further investigated by identifying the metabolites associated with either the dark or high light condition within the Chlorella fraction (Figure S3.4 and Table S3.1). This revealed a range of candidate symbiont metabolites that varied with light intensity. Metabolomics has an inherent trade-off between the confidence of the identifications and the extent the analysis is untargeted and unbiased. To take a truly untargeted approach with this many samples I did not use chromatography separation and therefore the identifications here are putative. The putative identifications associated with the shared dark response included amino acids and components of pyruvate and glycolysis metabolism. In addition, putative fatty acid and heme synthesis compounds were identified and these are known to be aspects of the non-photosynthetic roles of plastids (Barbrook et al., 2006), which suggests the Chlorella symbionts may fulfil additional functions, beyond carbohydrate supply, for their host. In contrast, the putative metabolites associated with the shared high-light response included plant hormones, a purine, carotenoids, carbohydrates and chlorophyll, indicating that photosynthesis and photoprotection characterised Chlorella metabolism in high light.

However, host-dependent differences in the metabolism of symbiont-genotypes could be detected. For the symbiont metabolite fraction, subset by host-genotype, I observed native versus non-native clustering of symbiont metabolism only when associated with the 186b host-genotype (PCA Figures 3.5, OPLS-DA Figure S3.3). This is consistent with the greater phenotypic differences in growth and symbiont load observed among host-symbiont pairings with the 186b host-genotype compared to with either the HK1 or HA1 host- genotypes.

Figure 3.4. The clustering of the metabolic fractions by light. These OPLS-DA plots show the metabolic fractions separated by light intensity, following clear PCA clustering by light (see PCA plots S3.2). Each point represents the metabolic profile of a sample; with the shape denoting the P. bursaria host genotype, the colour denoting the Chlorella symbiont genotype and the shade of the colour denoting the light intensity. Both fractions cluster according to shade, and therefore, according to light intensity. There are 3 replicates of each combination of host, symbiont and light intensity. A) The Chlorella metabolic fraction. B) The P . bursaria metabolic fraction.

Figure 3.5. Clustering patterns of the Chlorella metabolic fraction subset by host- genotype. These PCA plots show the HA1 host (A), the HK1 host (B), and the 186b host (C). Each point represents the metabolic profile of a sample; with the shape denoting the P. bursaria host genotype, the colour denoting the Chlorella symbiont genotype and the colour shade denoting the light intensity. Only within the 186b host (C) do the samples clusters by colour, and therefore, symbiont genotype. There are 3 replicates of each combination of host, symbiont and light intensity. 73

To identify the metabolites driving differences in metabolism of the symbiont-genotypes in the 186b host-genotype background, I next performed pairwise contrasts using volcano plots to highlight which metabolites varied significantly according to symbiont genotype (Figure 3.6). This revealed a range of candidate symbiont metabolites that varied between the native host-symbiont pairing and either of the non-native host-symbiont pairings (Table 3.1, 3.2 and 3.3). The putative identifications included, in the dark, elevated levels of candidate metabolites associated with stress responses (stress-associated hormones, jasmonic acid and abscisic acid, and stress associated-fatty acids, such as arachidonic acid) but reduced production of vitamins and co-factors by the native symbiont, compared to the non-native symbionts (Table 3.1). At high irradiance, the native symbiont showed higher levels of candidate metabolites in central metabolism, hydrocarbon metabolism and of biotin (vitamin B7), compared to the non-native symbionts (Table 3.3). In contrast, the non-native symbionts produced elevated levels, relative to native symbionts, of a candidate glutathione derivative, and glutathione is an antioxidant involved in the ascorbate- glutathione cycle that combats high UV stress through radical oxygen scavenging.

Figure 3.6. Differences in the Chlorella metabolism between symbiont genotypes at multiple light levels within the 186b P. bursaria host. Pairwise comparisons between symbiont genotypes are represented as volcano plots, plotting the fold change of each metabolite against its statistical significance. The data points are highlighted in red if the P value is significant and if the Log2(fold change) is greater than 1 or less than -1. Each panel compares two symbionts genotypes at one light level: A-C compare the 186b and HA1 symbionts, D-F compare the 186b and HK1 symbionts, and G-I compare the HA1 and HK1 symbionts. The first column is at the highest light level (50µE), the second column at the intermediate light level (12µE), and the third column is in the dark (0µE).

Table 3.1. Symbiont-genotype specific metabolites in the dark within the 186b P. bursaria host. These metabolite IDs were highlighted by the volcano plot (Figure 3.6) and were found to have significantly higher abundances in one symbiont-genotype compared to another within the 186b host subset of the Chlorella metabolic fraction in the dark (0µE). Recorded with 46ppm accuracy. Strain Detected Accurate Stress Associated Comparison mz ID Mass Mass Adduct Candidate Compound Pathway Associated s18 s18 vs sHA 118 117.966 117.0426 H+ Aspartate-4-semialdehyde Amino acid 117.0578 H+ Indole Amino acid + hormone 117.0790 H+ Glycinebetaine Amino acid + osmolyte 117.0790 H+ Valine Amino acid s18 vs sHA 134.2 134.109 133.1040 H+ Aspartate Amino acid s18 vs sHA 255.2 255.104 216.1725 K+ w-hydroxydodecanoic acid Hydroxy fatty acids 254.2246 H+ Palmitoleic acid Unsaturated fatty acids s18 vs sHA 343.2 343.153 342.1162 H+ Disaccaride Carbohydrate 304.2402 K+ Arachidonic acid Unsaturated fatty acids Yes 304.2402 K+ Kaurenoic acid Diterpenoid (related to GA) s18 vs sHK 247.2 247.117 224.1412 Na+ Methyl jasmonate Hormone (JA) Yes s18 vs sHK 267.2 267.102 228.2089 K+ Myristic acid Saturated fatty acids 244.2263 Na+ N1-acetylspermine Amino acid s18 vs sHK 271.2 271.167 248.1412 Na+ Abscisic acid aldehyde Hormone (ABA) Yes s18 vs sHK 686.4 686.391 663.3748 Na+ 1-Palmitoyl-2-(5-keto-6-octenedioyl)-sn- Glycerophospholipids Yes glycero-3-phosphocholine sHK s18 vs sHK 220.2 220.153 219.1107 H+ Pantothenate Vitamin (B5) 219.1120 H+ Zeatin Hormone (cytokinin) s18 vs sHK 238 238.053 199.0246 K+ O-phospho-L-homoserine Amino acid 215.0195 Na+ O-phospho-4-hydroxy-L-threonine Vitamin (B6) 215.0807 Na+ Kinetin Hormone (cytokinin) s18 vs sHK 241.2 241.188 202.2157 K+ Spermine Amino acid s18 vs sHK 335.2 335.115 334.2144 H+ Prostaglandin Fatty acyls 312.3028 Na+ Eicosanoic acid Saturated fatty acids sHK sHA vs sHK 355 355.048 354.0577 H+ 5-amino-6-(5'-phosphoribosylamino)uracil Riboflavin

Table 3.2. Symbiont-genotype specific metabolites in the intermediate light within the 186b P. bursaria host. These metabolite IDs were highlighted by the volcano plot (Figure 3.6) and were found to have significantly higher abundances in one symbiont-genotype compared to another within the 186b host subset of the Chlorella metabolic fraction in the intermediate light condition (12µE). Recorded with 46ppm accuracy.

Strain Detected Exact Stress Associated Comparison mz ID Mass Mass Adduct Candidate Compound Pathway Associated s18 vs sHA 361.4 361.303 338.3185 Na+ Erucic acid Fatty acid s18 s18 vs sHK 263.2 263.179 224.1412 K+ Methyl jasmonate Hormone Yes sHK s18 vs sHK 241.2 241.188 202.2157 K+ Spermine Amino acid s18 vs sHK 417.4 417.316 416.3654 H+ 6-oxocampestanol Hormone (Brassinosteroid) 417.4 417.316 416.3654 H+ Gamma-tocopherol Ubiquinone s18 vs sHK 451.2 451.12 450.1936 H+ Geranylgeranyl-PP Ubiquinone + Chlorophyll sHA sHA vs sHK 365 365.083 364.0420 H+ Xanthosine-5'-phosphate Purine 365 365.083 326.1226 K+ 6,7-dimethyl-8-(1-D-ribityl)lumazine Riboflavin

Table 3.3. Symbiont-genotype specific metabolites in the high light within the 186b P. bursaria host. These metabolite IDs were highlighted by the volcano plot (Figure 3.6) and were found to have significantly higher abundances in one symbiont-genotype compared to another within the 186b host subset of the Chlorella metabolic fraction in the highest light condition (50µE). Recorded with 46ppm accuracy. Strain mz Detected Accurate Stress Associated Comparison ID Mass Mass Adduct Compound Pathway Associated s18 s18 vs sHA 171 171.088 132.0059 K+ Oxalacetic acid TCA /central 169.9980 H+ Glycerone phosphate Glycolysis / central 169.9980 H+ Glyceraldehyde-3-phosphate Glycolysis / central 132.0423 K+ 3-hydroxy-3-methyl-2-oxobutanoate Amino acid 132.0423 K+ 2-acetolactate Amino acid 132.0423 K+ Glutarate Amino acid 132.0535 K+ Asparagine Amino acid

148.0372 Na+ Citramalate C5-Branched dibasic acid 132.0899 K+ Ornithine Amino acid 148.0736 Na+ Mevalonic acid Mevalonate pathway 148.0736 Na+ Pantoate Pantothenate biosynthesis s18 vs sHA 237.2 237.181 214.1317 Na+ Dethiobiotin Vitamin (B7) s18 vs sHA 239.2 239.145 200.1776 K+ Lauric acid Saturated fatty acids 216.1725 Na+ w-hydroxydodecanoic acid Hydroxy fatty acids s18 vs sHA 251.2 251.146 228.2089 Na+ Myristic acid Saturated fatty acids 212.2504 K+ Pentadecane Hydrocarbon s18 vs sHA 537.4 537.356 536.4382 H+ α/β/γ/δ carotene Carotenoid 536.4382 H+ Lycopene (all-trans or tetra cis) Carotenoid s18 vs sHK + sHA 213 213.097 174.0164 K+ Aconitic acid TCA cycle / central 190.0114 Na+ Oxalosuccinate TCA cycle / central 174.0528 K+ 3-Carboxy-4-methyl-2-oxopentanoate Amino acid 174.0528 K+ Shikimic acid Shikimate pathway 190.0477 Na+ 3-dehydroquinate Shikimate pathway 174.0793 K+ Indole-3-acetamide Amino acid + hormone 174.0892 K+ Suberic acid Fatty acid 174.1004 K+ N2-acetyl-L-ornithine Amino acid 190.1066 Na+ y-hydroxy-l-arginine Arginine-nitric oxide 212.0896 H+ Volemitol Carbohydrate 78

Table 3.3 continued Strain mz Detected Accurate Stress Associated Comparison ID mass mass Adduct Compound Pathway Associated s18 vs sHK + sHA 257.2 257.123 256.2402 H+ palmitic acid saturated fatty acid 256.1172 H+ 2-(3-Carboxy-3-aminopropyl)-L-histidine unusual amino acid s18 vs sHK 235.2 235.131 212.2504 Na+ pentadecane Hydrocarbon - metabolite sHK s18 vs sHK 220.2 220.153 219.1120 H+ Zeatin Hormone 219.1107 H+ Pantothenate Vitamin B5 sHK + sHA s18 vs sHK + sHK 465 465.096 426.0879 K+ S-Glutathionyl-L-cysteine Cysteine + methionine Yes sHK sHA vs sHK 329.2 329.1783 328.2402 H+ Docosahexaenoic acid Unsaturated fatty acids

Figure 3.7. Relative abundances of dark-stress associated metabolites across host genotypes in the dark. The data points show the mean (n=3) relative abundance ± SE. Each panel shows the data for one metabolite, with the colour distinguishing the symbiont-genotype. The three metabolites were associated with a prolonged darkness stress response for the native symbiont within the 186b host and are listed in Table 3.1. A) The mz bin 247.2, candidate compound: methyl jasmonate. B) The mz bin 271.2, candidate compound: Abscisic acid aldehyde. C) The mz bin 686.4.2, candidate compound: a Glycerophospholipid.

To test whether the 186b Chlorella underwent similar stress-responses when in the other host-genotype backgrounds, I next examined levels of the identified stress-associated candidate metabolites for 186b Chlorella across all host-genotypes. For dark-associated candidate stress metabolites, higher abundances were observed for 186b Chlorella in the 186b host-genotype background than in the HA1 or HK1 host-genotype backgrounds (Figure 3.7). These metabolite abundances were similar across all symbiont-genotypes in the HA1 or HK1 host-genotype backgrounds, suggesting that the dark-associated stress response of the 186b Chlorella is limited to its native host background and that dark- associated algal symbiont stress was ameliorated by the other host genotypes. In addition, the uniformity of the high-light stress response was tested by examining the abundance of the high-light candidate stress metabolites for the HA1 and HK1 Chlorella across all host- genotypes. As a group, these high-light stress associated metabolites did not have an overall clear pattern, although one metabolite had high abundances in the HA1 and HK1 Chlorella across all the host-genotype backgrounds (Figure 3.8). This implies that the high-light stress may therefore be independent of host-genotype.

Figure 3.8. Relative abundances of a high-light stress associated metabolite across host genotypes and across light levels. The data points show the mean (n=3) relative abundance ± SE. Each panel shows the data for a light level (0, 12 or 50 µE m-2 s-1) with the P. bursaria host-genotype on the X-axis and the symbiont-genotype shown by the colour. The metabolite is mz 465, candidate compound: S-Glutathionyl-L-cysteine (see Table 3.3). The metabolite was identified associated with a high-light stress response for the HA1 and HK1 symbionts within the 186b host.

3.4 Discussion

In this chapter, I investigated genetic variation for host-symbiont specificity in the P. bursaria - Chlorella endosymbiosis using a reciprocal cross-infection experiment coupled with metabolomics. I observed a significant GH × GS × E interaction for the host-symbiont growth rate that was predominately driven by the differential effects of symbiont-genotypes on host-symbiont growth rate within the 186b host, while in the other host-genotype backgrounds symbiont genotype did not affect growth rate. The regulation of symbiont load also displayed a GH × GS × E interaction driven by symbiont-genotype-specific responses in the 186b and HK1 host-genotypes. Consistent with the phenotype data, the metabolic profile of the Chlorella fraction varied when isolated from pairings with the 186b host genotype, but not with the other host-genotypes. The metabolic differences between symbionts in the 186b host potentially provides the mechanistic basis for the GH × GS × E interaction, and suggests that contrasting stress responses played an important role and

81 may have altered the benefit-to-cost ratio of symbiosis for this host. Specifically, whereas the 186b Chlorella showed a dark-associated stress response, producing stress-associated hormones and fatty acids, the HA1 and HK1 Chlorella showed a high-light-associated stress response, producing compounds to combat radical oxygen species. These data suggest that differences in light management among algal symbionts may underlie host-symbiont specificity, with implications for the likely success of partner switching.

The GH × GS × E interaction in the host-symbiont growth reaction norm reveals a striking asymmetry in specialisation among host genotypes. The growth rate reaction norm varied by symbiont genotype for the 186b host genotype, whereby the native symbiont was costlier in the dark but more beneficial in the high light environment compared to the non- native symbionts. This shows that the symbiont genotype affected the interaction between the benefit-to-cost ratio and light within the 186b host. In contrast, within the HK1 and HA1 host-backgrounds host-symbiont growth rate reaction norms did not vary according to symbiont genotype. Thus, whereas the HA1 and HK1 host-genotypes appear to be symbiont generalists, the performance of the 186b host genotype is far more dependent upon the genetic identity of its algal symbiont. The native 186b host-symbiont pairing appears to be specialised to high-light environments, showing high performance only within a limited range of high irradiances. Light specialism is common among photosynthetic organisms, such as the light ecotypes of the cyanobacteria Prochlorococcus (Rocap et al., 2003) and green alga Ostreococcus (Rodríguez et al., 2005), but this is the first time it has been shown in the P. bursaria – Chlorella endosymbiosis. The variation in host specialisation has implications for symbiont replacement via partner switching. Generalist hosts are likely to be more able to integrate novel symbiont-genotypes than more specialist hosts. Conversely, because diverse symbiont-genotypes result in similar growth reaction norms in generalist host genotypes, these host genotypes are probably less able to shift their ecological niche through partner switching. Nonetheless, variation in specialisation suggests that host-genotypes may vary extensively in the immediate fitness consequences of partner switching.

A previous mathematical model of the P. bursaria - Chlorella interaction suggested that symbiont load is host controlled and regulated to maximise the benefit-to-cost ratio of symbiosis (Dean et al., 2016; Lowe et al., 2016), offering a framework to understand the observed variation in the symbiont load reaction norms. Host regulation is believed to alter symbiont load through altering the balance of symbiont division/ingestion and symbiont digestion/egestion (Kodama and Fujishima, 2012; Takahashi et al., 2007). I observed a GH

× GS × E interaction in symbiont load data that is consistent with patterns of genotype- specific symbiont loads measured in other symbioses (Chong and Moran, 2016; Kondo et al., 2005), and is the first-time genotype-specific symbiont loads have been observed for the P. bursaria - Chlorella symbiosis. Whereas the HA1 host genotype regulated all symbiont genotypes in a similar manner, regulation varied according to symbiont genotype in the 186b and HK1 host-genotype backgrounds. In the HK1 host-genotype background the magnitude of the symbiont load altered according to symbiont genotype and the native symbiont had the highest symbiont load throughout ̶‒ a ‘native advantage’ outcome. In the 186b host-genotype background, the light intensity of the maximal symbiont load altered with symbiont genotype, specifically the symbiont load of the native symbiont peaked at a higher light intensity than the non-native symbionts. A change in the irradiance level at which the symbiont load curve peaks suggests fundamental differences in the benefit-to-cost ratio of these symbiont genotypes in response to light. The maximal symbiont load represents the point at which the benefit of the symbionts outweighs their cost. Symbiont loads that peak at a high irradiance imply that greater irradiance is required for the benefit per-symbiont to outweigh its cost (Dean et al., 2016). After the maximum load, symbiont load decreases with increasing irradiance because the energetic output per symbiont increases (Hoogenboom et al., 2010), and as such fewer symbiont are required to meet the demand (Dean et al., 2016; Lowe et al., 2016).

These phenotypic responses can be compared to the potential outcomes discussed in the chapter introduction (Figure 3.1). Within the HA1 host, symbiont genotype had no effect on growth rate or symbiont load, and therefore, HA1 appears to be a partner-generalist (similar to Figure 3.1a). Within the HK1 host the results were mixed; for growth rate, symbiont genotype had no effect, but symbiont load displayed a ‘native advantage’ outcome (similar to Figure 3.1b). A higher symbiont load, however, is not necessarily an advantage, and the discrepancy between the unaffected growth rate and increased symbiont load implies that the HK1 native symbiont is both less beneficial and less costly than the non-native symbionts. This is because the higher number of symbionts led to the same growth rate, suggesting that both the benefit and cost of symbioses was affected, but not the relationship between benefit and cost. In contrast, within the 186b host the growth rate and symbiont load depended on the interaction between symbiont genotype and the environment (similar to Figure 3.1c). Such that the relationship between the benefit-to- cost ratio and light differed according to symbiont-genotype, and this drives the GH × GS × E interaction.

The metabolomics data suggested that the native versus non-native symbiont-genotypes displayed contrasting stress-responses when inhabiting the 186b host-genotype background. In the dark, the 186b native symbiont-genotype had multiple candidate stress-response indicators, including stress-associated hormones and fatty acids. Prolonged darkness can trigger a stress response because the absence of photosynthesis can starve photosynthetic organisms of both fixed carbon and energy (Zhang et al., 2007); in plants and algae it has been demonstrated that the starch reserves are almost exhausted after one night (Graf et al., 2010; Ral et al., 2006). Starvation stress-responses are coordinated by signalling hormones and lipids such as those I have identified, especially abscisic acid (Lu et al., 2014), and arachidonic acid (Merzlyak et al., 2007). These signals can lead to downstream effects that try to negate the stress by mobilising alternative compounds for energy/carbon (Manoharan et al., 1999) or by triggering a resting state where metabolism and growth are reduced (Peters, 1996). The starvation stress response suggests that the symbiotic nutrient exchange has broken down within the 186b native symbiosis, and that neither partner is provisioning the other adequately. Consistent with these patterns, the 186b symbiont was costly and its load tightly regulated by the host in the dark when it was most stressed. Interestingly, in the other host backgrounds the 186b native symbiont did not display elevated dark-stress associated metabolites, and was not costly in these backgrounds. This suggests that the other hosts were perhaps more generous in provisioning their symbionts in the dark, and so prevented the starvation-based dark-associated stress response. Greater symbiont compatibility appears, therefore, partially due to the ability of hosts to prevent and ameliorate the stress response, and therefore, the cost of their symbionts.

In contrast, the HA1 and HK1 symbiont-genotypes in the 186b host-genotype background did not show dark-associated stress-responses, which, I hypothesise, is connected to their higher levels of candidate vitamins and cofactors that may help to stabilise cellular metabolism and therefore delay or prevent a full stress response (Abdel-Rahman et al., 2005; Asensi-Fabado and Munné-Bosch, 2010). At high irradiance, however, the pattern of stress-responses was reversed. Whilst the native 186b symbiont-genotype had no indicators of stress, the HA1 and HK1 non-native symbiont-genotypes showed indicators of high- light-mediated stress. Specifically, the HA1 and HK1 symbiont genotypes showed elevated levels of a candidate glutathione derivative; glutathione is an antioxidant involved in the ascorbate-glutathione cycle that scavenges reactive oxygen species to counteract the damaging consequences of excess light (Mallick, 2004; Shiu and Lee, 2005). It is well documented that increased antioxidant production is indicative of increased oxidative damage due to thermal or light stress (Bartosz, 1997; Lesser, 2006); in particular, the

84 consequences of this in symbiosis have been studied in coral, where oxidative stress causes the breakdown of symbiosis and coral bleaching (Lesser, 2011). The abundance profile of this candidate antioxidant revealed the HK1 and HA1 symbionts had high abundances across the host-backgrounds. Implying that while the dark-stress response is dependent on an interaction between symbiont-genotype and host-genotype, the high-light stress response is primarily dependent on symbiont-genotype. The consequences for host- symbiont growth vary between the symbiont genotypes, with the HA1 symbiont apparently capable of counteracting the stress without limiting growth rate, whereas for the HK1 symbiont stress from the high light intensity becomes excessive and its host-symbiont growth plateaus. This is supported by the candidate unsaturated fatty acid that had higher abundances in the HK1 symbiont compared to the HA1 symbiont in the high light environment. Unsaturated fatty acids are typically associated with higher stress (Klyachko‐ Gurvich et al., 1999; Thompson, 1996). Genetic variation in stress tolerance is observed in multiple photosymbioses, for example Symbiodinium-genotypes have different tolerance levels to high temperature stress (Cunning et al., 2015; Howells et al., 2012).

Surprisingly, I did not observe symbiont-genotype effects on the host metabolism. The absence of detectable genotype separation within the P. bursaria fraction could be a result of the differences being subtler than those of the Chlorella metabolism or alternatively due to biases in our detection methodology. Untargeted metabolomics attempts to be as unbiased as possible, but nonetheless extraction methods and machine settings will bias detection to metabolites with certain physiochemical properties (Ortmayr et al., 2016). For instance, there is often a trade-off between recording polar versus nonpolar metabolites and between sensitivity and dalton range of detection. Currently, the lack of host metabolic genotype separation precludes the investigation of how the HA1 and HK1 hosts compensates the 186b symbiont’s costliness in the dark and derive similar benefits of symbiosis from the different symbiont genotypes. This compensation could take the form of a metabolite that helps to ameliorate the stress response in the dark, or could be increased nutrient transfer to prevent starvation, perhaps the amino acid nitrogen compound which also contains fixed carbon. If the latter is true, then a pulse-chase metabolic experiment may be required to measure the transfer rates. This question remains open and hopefully future work will be able to investigate P. bursaria metabolomics in detail and untangle the host side of this interaction.

Stress is known to lead to the breakdown of symbioses (Abrego et al., 2008; Weis, 2008) and occurs in numerous environmental conditions, for example coral bleaching can be

85 caused by the stress of high temperature, high irradiance, prolonged darkness, or chemical pollution (DeSalvo et al., 2012; Douglas, 2003). In the Hydra – Chlorella endosymbiosis, the older, more stable origin of this association possesses greater oxidative stress tolerance compared to the more recent origin (Ishikawa et al., 2016). Stress tolerance and stress prevention are therefore likely to be crucial aspects of symbioses, and particularly so in photosymbioses that cannot escape the potentially damaging consequences of light. This chapter has shown how contrasting light-dependent symbiont stress-responses drive host- symbiont genetic specificity by altering the benefit-to-cost ratio of the symbiosis. A result that corresponds with the role of symbiont-stress tolerance in determining the performance of other photosymbioses (Abrego et al., 2008; Howells et al., 2012; Ye et al., 2019), providing evidence that this interaction may be a common feature of photosymbioses. Furthermore, the comparison of stress metabolites across the host-genotypes suggested that generalist host profiles occur in genotypes that alleviate stress in their partners, leading to similar benefit-to-cost ratios across their symbionts. The alleviation of stress, therefore, may not only affect the fitness outcome of a symbiosis but also the likelihood of novel symbiont integration, and thus partner switching. It would be interesting to examine whether this holds true for other photosymbioses, as the role of stress in partner integration could be consequence of genotype-driven stress tolerance playing such a pivotal role in photosymbioses. Photosymbioses are inherently tied to photo-oxidative stress, and it appears that the evolution of these endosymbioses is driven by their adaptation to, and tolerances of, stress.

3.5 Supplementary Figures

186b host HA1 host HK1 host L s18 sHA sHK h18 hHA hHK k18 kHA kHK

1000 bp

500 bp

Figure S3.1. PCR confirmation of symbiont-genotype within the reciprocal cross infections. Overlapping, multiplex primers were used to amplify fragments within the 18S rDNA and ITS region of the Chlorella nuclear genome. In this region the ‘American/Japanese’ strains, such as HA1 and HK1, have had three introns inserted that the ‘European’ strains, such as 186b, lack (Hoshina and Imamura, 2008; Hoshina et al., 2005). The main fragment of HA1/HK1 is, therefore, considerably larger than the main fragments of 186b, and both have additional smaller fragments. The banding pattern results here confirm that the cross- infections were successful and contain the correct Chlorella genotype, specifically that the distinct banding pattern of 186b was present when expected. This PCR method can distinguish between ‘American/Japanese’ and ‘European’ strains, but not between strains that come from the same biogeographical clade. Host genotype has been shortened to a letter (‘s’ = 186b host, ‘h’ = HA1 host, ‘k’ = HK1 host); symbiont genotype is shown by two capitals (‘18’ = 186b symbiont, ‘HA’ = HA1 symbiont, ‘HK’ = HK1 symbiont. Shown alongside a 100bp ladder.

Figure S3.2. The clustering of the metabolic fractions by light in PCA plots. These PCA plots show the initial clustering of the metabolic fractions by light, which was then shown by OPLS-DA plots (Figure 3.3). Each point represents the metabolic profile of a sample; with the shape denoting the P. bursaria host genotype, the colour denoting the Chlorella symbiont genotype and the shade of the colour denoting the light intensity. There are 3 replicates of each combination of host, symbiont and light intensity. A) The Chlorella metabolic fraction. B) The P. bursaria metabolic fraction.

Figure S3.3. Separation by symbiont-genotype within the 186b host subset of the Chlorella metabolic fraction. This OPLS-DA plot follows the initial clustering by symbiont- genotype within the 186b host subset in the PCA plot (Figure 3.4C). Each point represents the metabolic profile of a sample; with the colour denoting the Chlorella symbiont genotype and the shade of the colour denoting the light intensity. The samples separate between the ‘blue’ samples (186b symbiont-genotype) and the ‘green’ and ‘grey’ samples (HA1 and HK1 symbiont genotypes). There are 3 replicates of each combination of host, symbiont and light intensity.

Figure S3.4. Shared response of Chlorella genotypes to light intensity in the Chlorella metabolic fraction. Pairwise comparison between the dark (L0 = 0µE) and high light level (L50 = 50µE) across genotypes represented as a volcano plot, plotting the fold change of each metabolite against its statistical significance. The data includes Chlorella data from all nine of the cross-infections, and therefore indicates the shared response, irrespective of host or symbiont genotype. The data points are highlighted at two false discovery rate (FDR) values, and if the Log2(fold change) is greater than 1 or less than -1.

3.6 Supplementary Tables Table S3.1. Light-level associated shared Chlorella metabolites across the host and symbiont genotypes. These metabolite IDs were identified from the top compounds highlighted in the volcano plot Figure S3.4. They have, therefore, statistically significantly higher abundances in either the dark or high light, within the Chlorella metabolic fraction. Light mz Detected Accurate Accurate Association ID Mass + Adduct Mass Adduct Candidate Compound KEGG ID Pathway 0µE 69 69.031 68.974 30.011 K+ Formaldehyde C00067 Methane 69 69.031 68.983 45.993 Na+ Nitrite C00088 Nitrogen 69 69.031 68.995 46.005 Na+ Formate C00058 Pyruvate + Methane 69 69.031 69.032 46.042 Na+ Ethanol C00469 Glycolysis 73 73.009 73.02838 72.021 H+ Methylglyoxal C00546 Pyruvate 75 75.023 75.008 74.000 H+ Glyoxylic acid C00048 Central (Glyoxylate cycle) 75 75.023 75.045 74.037 H+ Lactaldehyde C00424 Carbohydrate + Pyruvate 75 75.023 75.045 74.037 H+ Propanoic acid C00163 Propanoate (lipid) 75 75.023 75.044 74.037 H+ Hydroxyacetone C05235 Propanoate (lipid) 105 105.033 105.019 104.011 H+ Hydroxypyruvic acid C00168 Amino acid + Photorespiration 105 105.033 105.019 104.011 H+ Malonate C00383 Fatty acid 105 105.033 105.030 104.022 H+ Urea-1-carboxylate C01010 Urea degradation 154 153.993 154.011 131.022 Na+ Iminoaspartate C05840 Nicotinate 154 153.993 154.027 115.063 K+ Proline C00148 Amino acid 154 153.993 154.048 131.058 Na+ 5-Aminolevulinate C00430 Heme Biosynthesis 154 153.993 154.048 131.058 Na+ Glutamate-5-Semialdehyde C01165 Amino acid 154 153.993 154.048 131.058 Na+ 4-Hydroxy-proline C01157 Amino acid 154 153.993 154.050 153.043 H+ 3-Hydroxyanthranilate C00632 Amino acid 154 153.993 154.059 131.069 Na+ Creatine C00300 Amino acid 154 153.993 154.084 131.095 Na+ B-Alaninebetaine C08263 Osmoprotectant/stress 154 153.993 154.084 131.095 Na+ Isoleucine C00407 Amino acid + Cyanoamino 154 153.993 154.084 131.095 Na+ Leucine C00123 Amino acid 154 153.993 154.087 153.079 H+ Dopamine C03758 Alkaloid + Amino acid 154 153.993 154.096 131.106 Na+ N-Carbamoylputrescine C00436 Amino acid

Table S3.1 continued Light mz Detected Accurate Accurate Association ID Mass + Adduct Mass Adduct Candidate Compound KEGG ID Pathway 0µE 212 212.022 212.033 173.069 K+ N-Acetyl-glutamate-semialdehyde C01250 Amino acid 212 212.022 212.053 189.064 Na+ N-Acetyl-glutamate C00624 Amino acid 425.2 425.176 425.100 386.137 K+ Pteryxin C09307 Coumarins 425.2 425.176 425.100 386.137 K+ Samidin C09310 Coumarins 425.2 425.176 425.135 424.127 H+ Adifoline C09020 Indole alkaloid 521.4 521.386 521.311 520.304 H+ Cyasterone C08816 Sterol Lipid + Terpenoid 651.2 651.241 651.196 612.232 K+ Novobiocin C05080 Antibiotic 50µE 242.2 242.193 242.100 219.111 Na+ Pantothenate C00864 Pantothenate + CoA 242.2 242.193 242.102 219.112 Na+ Zeatin C15545/C00371 Plant hormone 242.2 242.193 242.125 241.118 H+ Tetrahydrobiopterin C00272 Folate biosynthesis 300.2 300.186 300.160 299.152 H+ Codeine C06174 Isoquinoline alkaloid 300.2 300.186 300.107 277.118 Na+ Queuine C01449 Nucleobase + Purine 365 365.08 365.050 364.042 H+ Xanthosine-5-phosphate C00655 Purine 365 365.08 365.085 326.121 K+ Neohesperidose C08244 Carbohydrate 365 365.08 365.085 326.121 K+ Robinobiose C08246 Carbohydrate 365 365.08 365.086 326.123 K+ 6,7-dimethyl-8-(D-ribityl)lumazine C04332 Riboflavin 448.2 448.117 448.122 409.158 K+ Linustatin C08333 Cyanogenic glucosides 531.4 531.367 531.454 492.491 K+ Tritriacontane-16,18-dione C08394 Alkane 569.4 569.338 569.313 568.305 H+ Protoporphyrinogen IX C01079 Chlorophyll 569.4 569.338 569.436 568.428 H+ Xanthrophyll C08601 Carotenoid 569.4 569.338 569.436 568.428 H+ Zeaxanthin C06098 Carotenoid + Hormone (ABA) 585.4 585.329 585.431 584.423 H+ Antheraxanthin C08570 Carotenoid 664.2 664.196 664.267 625.303 K+ Leukotriene C4 C02166 Lipid - Arachidonic acid

Chapter 4

A novel host-symbiont interaction can rapidly evolve to become a beneficial symbiosis

4.1 Introduction

Endosymbioses are evolutionarily dynamic. Their environmental context dependence (Thompson, 2005) generates inherent potential for conflicting fitness interests among the symbiotic partners (Sachs and Simms, 2006). This can lead to the breakdown of symbiosis if environmental conditions change faster than symbionts can adapt or where pursuit of individual fitness interests lead to the emergence of cheating. Both situations can create selection for partner switching to recombine novel symbiotic partnerships (Boulotte et al., 2016). Partner switching can restore symbiont function following breakdown (Koga and Moran, 2014; Matsuura et al., 2018) or where the current symbiotic phenotype is maladapted to prevailing environmental context (Lefèvre et al., 2004). As such, partner- switching can enable niche-expansion by hosts (Joy, 2013; Sudakaran et al., 2017) and provide a mechanism by which host-symbiont local adaptation can arise faster than by adaptation of the current symbiont (Jaenike et al., 2010; Jiggins and Hurst, 2011). For example, corals acquire thermally-tolerant Symbiodinium endosymbionts following thermal bleaching events (Boulotte et al., 2016), and replacement of the photobiont with an alternative ecotype is believed to have enabled symbiont-mediated niche expansion in lichens (Rolshausen et al., 2018). A greater understanding of partner switching is also required if we are to understand life-history patterns; specifically, serial symbiont replacements have occurred in plastid evolution and have entangled the eukaryotic lineages (Patron et al., 2006; Stiller et al., 2014). The frequency of partner switching in natural populations suggests that new host-symbiont genotype pairings must arise regularly in a wide range of symbioses.

Despite the widespread occurrence of partner-switching, however, new host-symbiont pairings may have low fitness because the genotypes are unlikely to be co-adapted due to a lack of recent coevolutionary history. This has been observed in a range of symbiotic interactions: for example, a newly acquired Symbiodinium endosymbiont was found to

93 translocate less fixed carbon than the native symbiont to its cnidarian host (Matthews et al., 2018); novel bacterial endosymbionts had reduced vertical transmission rates in aphid hosts (Russell and Moran, 2005); and novel Wolbachia endosymbionts reduced the reproductive fitness of Drosophila simulans (McGraw et al., 2002). How then are these newly-formed, poorly co-adapted host-symbiont pairings stabilised? Experimental studies suggest that initially low fitness host-symbiont associations can rapidly ameliorate their initial fitness costs: For example, the higher fitness cost of novel Spiroplasma endosymbiont genotypes could be rapidly alleviated in Drosophila melanogaster (Nakayama et al., 2015). I hypothesised, therefore, that this process could be enabled by rapid evolution, to create a beneficial, co-adapted association from a low fitness starting point.

Experimental evolution provides a powerful tool to study the dynamics of symbiotic interactions as it allows evolutionary processes to be studied in real time and in controlled laboratory conditions (Hoang et al., 2016). Previous applications of experimental evolution to symbiosis have studied the evolution of entirely de novo associations (Jeon, 1987; Nakajima et al., 2009, 2015), as well as the transition between parasitism and mutualism (King et al., 2016; Sachs and Wilcox, 2006; Shapiro and Turner, 2018; Tso et al., 2018). Rarely, however, has experimental evolution been used to study beneficial endosymbioses (Hoang et al., 2016). To test the role of rapid evolution in the establishment of new host- symbiont associations, I recapitulated the process of partner-switching by creating a novel Paramecium bursaria - Chlorella association which had initially low fitness, and then experimentally evolving replicate populations. The P. bursaria - Chlorella symbiosis is primarily based on a nutrient exchange between fixed carbon from the photosynthetic alga and organic nitrogen from the heterotrophic host (Johnson, 2011; Ziesenisz et al., 1981). It is highly experimentally tractable and amenable to experimental evolution: symbiotic P. bursaria have fast generation times and replicate populations can easily be cultured in the laboratory. To my knowledge, the Paramecium bursaria - Chlorella endosymbiosis has not been used previously for experimental evolution, although an evolution experiment using another member of the Paramecium genus has been published. Lohse et al. (2006) coevolved Paramecium caudatum with a bacterial parasite for 130-260 host-generations, reporting that hosts evolved greater resistance against their coevolved parasites. In addition, Chlorella vulgaris was shown to rapidly adapt to predation within a few generations (Yoshida et al., 2004). These demonstrate that both Paramecium and Chlorella are capable of rapid evolutionary responses to selection.

I used the results from Chapter 3 to choose a newly-formed P. bursaria - Chlorella association that was less-beneficial than the native association. I chose the 186b host and HK1 symbiont pairing, which had lower growth rate than the native 186b pairing at high light (Figure 3.2). Furthermore, this novel association had lower symbiont load than the native pair at high light (Figure 3.3), which offered a potential mechanism for selection to act upon. I predicted, therefore, that the novel symbiosis would evolve upregulation of its symbiont load to increase the benefit to the host and so increase its growth rate. Replicate populations of the novel 186b-HK1 association were experimentally evolved by serial transfer for ~50 host generations and compared to control populations of the native association that were evolved under identical conditions. I tracked changes in host- symbiont growth rate and per host symbiont load over the course of the experiment, and quantified change in fitness between the start and end of the experiment. To determine the mechanisms of adaptation, and distinguish host versus symbiont contributions, I performed untargeted metabolomics separately on the host and symbiont at the start and end of the experiment. I observed that the initially non-beneficial novel host-symbiont association rapidly evolved to become as beneficial as the native host-symbiont pairing. The data further suggest that this was driven by changes in symbiont load and metabolism. These data confirm that rapid evolution can indeed enable non-beneficial host-symbiont pairings arising from partner-switching to become highly beneficial in fewer than 50 host generations.

4.2 Materials and Methods

Cultures & Strains P. bursaria - Chlorella cultures were maintained under the conditions described in Chapter 2. The two symbiotic partnerships used in this chapter were: the 186b host-genotype with its native symbiont and the 186b host-genotype with the non-native, HK1, symbiont. These were created from the cross-infections in Chapter 3.

Evolution Experiment The populations used derive from the cross-infections in Chapter 3, and, therefore, the symbiotic partnerships come from the same cured 186b ancestor that was then re-infected with either its native (186b) or novel (HK1) symbionts. The two symbiotic partnerships were split into six replicate populations that were used as the starting populations. The 200ml populations were propagated by weekly serial transfer for 25 transfers at a high light 95

(50 µE m-2 s-1) 14:10 L:D cycle. At every transfer, cell-density was equalised to 100 cells ml-1 and the transferred cells were washed with a 11µm nylon mesh using Volvic before being re-suspended in bacterized PPM. Cell density was measured before and after each transfer by fixing 360 µL of each cell culture, in triplicate, in 1% v/v glutaraldehyde in 96- well flat bottomed micro-well plates. Images were taken with a plate reader (Tecan Spark 10M) and cell counts were made using an automated image analysis macro in ImageJ v1.50i (Schneider et al., 2012). Fitness assays were conducted at the start and end of the experiment as described in Chapter 2. Growth rate and symbiont load assays were conducted at the start, T10, T20 and end of the experiment described in Chapter 3.

Metabolomics The cultures were sampled at the start and end of the evolution experiment. Cultures were washed and concentrated with a 11µm nylon mesh using Volvic and re-suspended in bacterized PPM. The cultures were acclimated at their treatment light condition (50 µE m-2 s-1) for seven days. For the start point, the six experimental replicates were used as replicates for the metabolomics. For the end point, three replicates of each of the six experimental replicates were used for the metabolomics because divergence may have occurred over the course of the experiment.

At each sampling event, the symbiotic partners were separated in order to a get P. bursaria and Chlorella metabolic fraction using the extraction method described in Chapter 2. Samples were freeze-dried for storage, and then resuspended in 50:50 methanol to water prior to mass spectrometry.

The samples were analysed with a Synapt G2-Si with Acuity UPLC, recording in positive mode over a large untargeted mass range (50 – 1000 Da). A 2.1x50mm Acuity UPLC BEH C18 column was used with acetylnitrile as the solvent. The machine settings are listed in detail below:

Gradient information: Time (mins) W ater (%) Acetonitrile (%) 0 95 5 3 65 35

6 0 100

7.5 0 100 7.6 95 5

The P. bursaria and Chlorella fraction were analysed separately. The xcms R package (Benton et al., 2010; Smith et al., 2006; Tautenhahn et al., 2008) was used to extract the spectra from the CDF data files, using a step argument of 0.01 m/z. Peaks were identified, and then grouped across samples. These aligned peaks were used to identify and correct correlated drifts in retention time from run to run. Pareto scaling was applied to the resulting intensity matrix.

Metabolomics Analysis The metabolic profiles from the start and end of the experiment were compared using principal component analysis (PCA) with the prcomp() function in Base R (https://www.r- project.org/). For both fractions the first three components were considered, this accounted for >88% of the variance. The top 1% of the loadings were selected using the absolute magnitude of the loadings. These top loadings were identified where possible, and the identified loadings were then depicted in their associated component space. The relative abundance of these top loadings was visualised using heatmaps drawn with the heatmap.2() function from the gplot package (Warnes et al., 2009). The phylogenies were based on UPGMA clustering of the PCA coordinates of the samples using the hclust() function.

Data Analysis Statistical analyses were performed in Rv.3.5.0 (R Core Team, 2018) and all plots were produced using package ggplot2 (Wickham, 2016) unless otherwise stated. Physiology tests

were analysed by both ANOVA and ANCOVA, with transfer time, host and symbiont identity as factors. A linear mixed effect model was used to analysis the growth rate per transfer using lm() function from the nlme package (Pinheiro et al., 2019). The lm model included fixed effects of symbiont genotype and transfer number, and random effects of transfer number given sample ID.

4.3 Results

Replicate experimental populations of either the novel host-symbiont pairing or the native host-symbiont pairing were established. Specifically, the 186b P. bursaria - Chlorella strain was cured of its native algal symbiont and subsequently re-infected with either its native algal symbiont or the novel HK1 algal symbiont. Six replicate populations of each of these two symbiotic partnerships were then propagated by weekly serial transfer for 25 transfers at a high light (50µE) 14:10 L:D regime. At every transfer cell-density was equalised to 100 cells ml-1 among populations to prevent extinctions. The growth rate per transfer was higher for the native pairing than the novel pairing (Figure 4.1) (linear mixed effect model, HK1 symbiont fixed effect of -0.08 ±0.006, T-value = -14.126, see Appendix D for full statistical output), but increased over time for both pairings (transfer number fixed effect 0.001 ±0.0004, T-value = 3.088).

Figure 4.1. Weekly growth rates of the native and novel symbioses across the evolution experiment. The lines show the smoothed mean (n=6) growth rates ± SE and colour denotes the symbiont genotype (blue = s18 = native symbiont; grey = sHK = novel symbiont). The smoothing function used was the loess method. 98

To test for change in symbiotic performance I quantified the host-symbiont growth rate reaction norm across a light gradient at multiple points during the experiment (Figure 4.2). At the beginning of the experiment, the growth rate of the native pairing increased more steeply with irradiance than the novel pairing, suggesting that the host derived greater symbiotic benefit at higher light intensity from its native symbiont. This difference was reduced over time, such that both the native and the novel pairings showed equivalent growth rate responses with light irradiance by the endpoint of the experiment (ANOVA,

F13,178 =56.14, P<0.001). This compensation for the initially poor performance of the novel symbiont at high irradiance occurred rapidly, such that the host-symbiont growth rate reaction norms of the native and novel pairings already appeared similar by transfer 10. These data suggest that newly established symbioses can rapidly achieve similar growth performance as the native host-symbiont pairing.

Figure 4.2. Growth rate assays performed at multiple points throughout the evolution experiment. Each panel shows the mean (n=6) initial growth rate across a light gradient and the shaded area denotes ± SE. The panels represent the transfer week within the evolution experiment at which the growth assay was performed (T0 = week 0, T10 = week 10, T20 = week 20 & T25 = week 25). The symbiont-genotype is denoted by colour.

In this symbiosis, hosts are known to regulate the cost-to-benefit ratio of symbiosis by altering symbiont load. To determine if regulation of symbiont load changed during the transfer experiment, symbiont load was measured as the intensity of single-cell fluorescence using a flow cytometer following growth across a light gradient (Figure 4.3). At the start of the experiment both host-symbiont pairings showed the expected unimodal symbiont load curve with light, albeit with higher symbiont loads for the native compared to the

99 novel pairing at the highest light level, 50 µE m-2 s-1, which was the irradiance used in the transfer experiment. By the end of the transfer experiment, the functional forms of the symbiont load reaction norms had changed in both host-symbiont pairings (symbiont genotype*light*transfer interaction, ANOVA, F19,76 = 34.15, P<0.001). Most notably, while the novel pairing had increased symbiont load at 50 µE m-2 s-1, symbiont load had decreased in the native pairing at this irradiance, such that symbiont load was now higher in the novel pairing at 50 µE m-2 s-1 irradiance. This increase in novel symbiont load at high irradiance may explain the improved growth performance observed at high light in the novel pairing. Interestingly, whilst the novel pairing retained the characteristic unimodal relationship between symbiont load and irradiance during the transfer experiment, this appears to have been lost in the native pairing, suggesting that altered symbiont load regulation can arise when evolved in a consistent light-dark environment.

Figure 4.3. Symbiont load at the start and end of the evolution experiment. Symbiont load was measured across a light gradient. The left-hand panel shows data measured at the start of the evolution experiment and the right-hand panel shows data measured at the end. The points show the mean (n=6) relative chlorophyll fluorescence ± SE and symbiont- genotype is denoted by colour.

100

To compare the fitness effect of symbiosis for the host before and after evolution, I directly competed the native and novel symbiotic pairings against the ancestral symbiont-free host strain across a light gradient at the beginning and end of the transfer experiment. Specifically, I used flow cytometry to quantify the proportion of symbiotic versus non- symbiotic cells at the start and end of competitive growth and calculated the selection rate, providing a direct measure of the fitness effects of symbiosis. At the beginning of the transfer experiment, the fitness of symbiotic relative to non-symbiotic hosts increased more steeply with irradiance for the native than the novel pairing (Figure 4.4). Following evolution, this difference had disappeared such that both the native and novel symbiotic pairings showed increasing fitness relative to non-symbiotic hosts with increasing irradiance -2 -1 (ANOVA, F11,41 =8.87, P<0.001). Indeed, at 50 µE m s , the light level used in the selection experiment, the large fitness deficit observed between the novel and native pairing at the beginning of the experiment had been completely compensated following evolution.

Figure 4.4. Fitness of the host-symbiont pairings relative to the symbiont-free host at the start and end of the evolution experiment. Lines show mean (n=6) competitive fitness of symbiont-containing hosts relative to the symbiont free 186b host calculated as selection rate, the shaded area denotes ± SE. The left-hand panel shows data measured at the start of the evolution experiment and the right-hand panel shows data measured at the end. Symbiont-genotype is denoted by colour. A selection rate above 0 indicates greater fitness in comparison to the symbiont-free host.

101

It was no longer possible to cure the evolved host-symbiont pairings of their symbionts, and so to estimate the contribution of host versus symbiont evolution to the observed convergence in host-symbiont fitness I used metabolomics. Specifically, I performed untargeted metabolomics analyses on the separated Chlorella and P. bursaria fractions from samples taken the start and end of the transfer experiment grown at 50 µE m-2 s-1. The ancestral P. bursaria and Chlorella metabolic profiles of both host-symbiont pairings could be clearly distinguished. Following evolution, P. bursaria metabolism displayed a high degree of convergence between hosts evolved with the native versus the novel symbionts (Figure 4.5 a,c). This was driven by decreased levels of compounds of central metabolism (such as pyruvate and TCA cycle intermediates, antioxidants, lipids, and some amino acids) (Table S4.1), suggesting either increased pathway completion or a reduced metabolic rate, both of which can lead to increased efficiency. In addition, there were increased levels of the amino acid cysteine and a shikimate pathway component (Figure 4.6). I also observed increased levels of algae-cell degradation components, such as cell-wall degradation product chitotriose, in some replicates with either symbiont, potentially suggesting increased digestion of Chlorella (Figure 4.6). In contrast, the metabolic profiles of the symbiont genotypes were less consistent (Figure 4.5 b,d). Whereas all replicates of the native 186b Chlorella evolved in a similar direction, the replicates of the novel HK1 Chlorella split into two different directions. Two of the HK1 replicates took a similar trajectory to the 186b symbionts, while the remaining four replicates followed an opposing evolutionary trajectory. The group of four HK1 replicates that diverged during the experiment, had lower production of metabolites within core aspects of metabolism, such as lipids, amino acids and carbohydrates. The second group including the remaining two HK1 replicates and all the 186b replicates had higher production within primary metabolism pathways, particularly within lipids and carbohydrates, as well as a key chlorophyll compound, a photo-protective carotenoid, and secondary metabolites with potential antioxidant properties (Figure 4.7, Table S4.2). This greater investment into photosynthesis and photo-protection may improve carbon transfer to the host and decrease light stress, which aligns with the decrease in host antioxidants.

102

Figure 4.5. The trajectories of the metabolic profiles from the start to the end of the evolution experiment. These trajectories are shown within PCA plots and the arrows represent the movement in principal component space over the course of the experiment, with 95% confidence ellipses drawn for the evolved profiles. The metabolite identifications for the top loadings are shown in their corresponding location. Colour denotes the symbiont-genotype and shade represents whether the samples are from the start or end of the experiment. A and C show the results for the P. bursaria fraction, B and D the Chlorella fraction. The top row (A and B) plot PCA 1 versus PCA 2. The bottom row (C and D) plot PCA 2 versus PCA 3. The data here presents the biological replicates, which have been averaged over their technical replicates.

103

Figure 4.6. Metabolites of interest across the start and end of the evolution experiment within the P. bursaria fraction. The data is shown as a heatmap with the colour representing the relative abundance of the metabolites. The metabolites depicted were identified from the top loadings of the PCA plots. The columns in the heatmap correspond to sample; these are labelled with their symbiont-genotype (18 = 186b; HK = HK1) and with the replicate number if from the end of the experiment or ‘start’ if from the start. The column on the left of the heatmap indicates the function of the metabolites and the column on the right indicates the loading ID, which corresponds to the identification Table S4.1. The phylogeny of the samples was calculated with their principal component coordinates using UPGMA clustering, and the order of the rows was assigned by UPGMA clustering performed on the rows’ distance measures (based on the Pearson correlation co-efficient).

104

Figure 4.7. Metabolites of interest across the start and end of the evolution experiment within the Chlorella fraction. The data is shown as a heatmap with the colour representing the relative abundance of the metabolites. The metabolites depicted were identified from the top loadings of the PCA plots. The columns in the heatmap correspond to the sample; these are labelled with their symbiont-genotype (18 = 186b; HK = HK1) and with the replicate number if from the end of the experiment or ‘start’ if from the start. The column on the left of the heatmap indicates the function of the metabolites and the column on the right indicates the loading ID, which corresponds to the identification Table S4.2. The phylogeny of the samples was calculated on their principal component coordinates using UPGMA clustering, and the order of rows was assigned by UPGMA clustering performed on the rows’ distance measures (based on the Pearson correlation co-efficient).

4.4 Discussion

In this chapter, I show that an initially non-beneficial, novel host-symbiont pairing evolved to become as beneficial as the native host-symbiont pairing in fewer than 50 host generations. This increase in the fitness benefit of symbiosis to hosts was accompanied by increased symbiont load following evolution in the novel, but not the native, host-symbiont pairing at the irradiance level in which they had evolved. I observed convergence of P. bursaria metabolism between the native and novel host-symbiont pairings following

105 evolution. Specifically, decreased levels of the intermediates of central metabolism, antioxidants and lipids, were observed following evolution compared to samples from the start of the experiment. Multiple trajectories of metabolic evolution were observed in Chlorella among the replicates of the novel host-symbiont pairing, two of which increased their investment into metabolites associated with photosynthesis and photo-protection. Together these data suggest that newly-formed host-symbiont pairings can rapidly evolve higher fitness through changes in symbiont load regulation and metabolism.

Consistent with findings from other symbioses (Matthews et al., 2018; Nakayama et al., 2015), I observed initially low fitness of the newly-formed host-symbiont pairing, suggesting poor co-adaptation arising from a lack of coevolutionary history. This initial lack of co- adaptation is likely a consequence of the novel symbiont belonging to a different clade than the native symbiont, and therefore these symbiont genotypes arise from independent origins of this symbiosis (Hoshina and Imamura, 2008; Summerer et al., 2008). The Chlorella clades are biogeographical and it is highly unlikely these symbiont-genotypes co- occur (Hoshina et al., 2005) and, therefore, the partners of the novel association have probably never encountered one another before. Furthermore, the previous chapters of this thesis have shown that the symbiont clades are associated with different photophysiology traits and stress tolerances that affect the light-dependent fitness outcome of this association. Trait variation between the clades could explain the initially mis-matched symbiotic phenotype that caused low fitness of the novel host-symbiont pairing.

Following evolution, the initially low fitness novel host-symbiont pairing acquired high fitness benefits equivalent to the native host-symbiont association. Phylogenetic reconstruction has predicted that many beneficial bacterial symbionts originated as parasites (Sachs et al., 2011), which demonstrates that the evolution of benefit also occurs in nature and can led to stable mutualisms. In addition, the evolution of benefit has been documented within evolution experiments (King et al., 2016; Nakayama et al., 2015; Shapiro and Turner, 2018; Tso et al., 2018). My findings, therefore, align with previous results and extend the evidence to include the evolution of benefit from a novel, non- beneficial symbiosis.

The evolution of increased benefit within the novel association was partially mediated by increased symbiont load, which could directly affect the benefit-to-cost ratio of the symbiosis (Holland et al., 2002, 2004). Corals are known to adjust the per host load of Symbiodinium endosymbionts according to environmental conditions and symbiont

106 genotype to maximise the benefit of the symbiosis (Cunning et al., 2015). Similar regulation of symbiont load with light intensity has been observed in the P. bursaria - Chlorella association (Dean et al., 2016; Lowe et al., 2016). Such that the symbiont load peaks in the low light, when the benefit of the symbionts outweighs their cost, and then the symbiont load decreases with increasing irradiance as the benefit-per-symbiont increases (Hoogenboom et al., 2010) and fewer symbionts are required to meet the demand. Changes to symbiont load can occur by hosts either triggering symbiont cell division or by the digestion/egestion of symbionts (Kodama and Fujishima, 2008; Takahashi et al., 2007). The evolution of increased symbiont load in the novel host-symbiont pairing implies that, initially, hosts had too few symbionts to meet their demand for fixed carbon. Accordingly hosts upregulated symbiont load during the course of the experiment, leading to a higher host-symbiont growth rate. In contrast, in the control native host-symbiont pairing, symbiont load at the light level at which the populations had evolved declined following evolution without any reduction in host-symbiont growth rate. This implies that the benefit-per-symbiont accrued to the host may have increased during evolution in the native but not in the newly formed association, suggesting that evolution can further fine-tune even established host-symbiont associations.

The difference between the evolved symbiont loads may be explained by the metabolic adaptation of the symbionts, which affects their benefit to the host. The native symbiont evolved higher production of several key metabolites, including chlorophyll, lipids and carbohydrates. These metabolites imply that increased symbiont investment into photosynthesis may have increased the carbon transferred to the host; a correlation that has been observed in other photosymbioses (Cantin et al., 2009; Freeman et al., 2013). In turn, the increased translocated carbon could have driven the increased benefit-per- symbiont and, therefore, led to the observed reduced symbiont load of the native symbiont. On the other hand, the metabolic profiles of the novel symbiont evolved in two directions. Two of the HK1 replicates converged with the 186b symbionts, having higher production of the same key metabolites. The remaining HK1 replicates, however, did not evolve increased photosynthetic investment or carbon transfer but decreased the metabolic intensity of some of these key compounds, and therefore did not converge with the metabolism of the native symbiont. To test whether the alternative metabolic trajectories affected the benefit of the symbionts, I examined the per-replicate symbiont load, on the premise that a higher symbiont load should be associated with a lower benefit-per- symbiont. Within this data (Table S3), the two HK1 replicates that converged metabolically with the native symbionts had a lower increase in symbiont load compared

107 to the replicates that metabolically diverged. This implies that while a few of the novel symbionts evolved in a similar manner to the native symbionts and increased their benefit to the host, the majority of the novel symbiont replicates did not.

Over the course of the experiment, the P. bursaria metabolism converged between hosts harbouring the native and novel symbiont. The convergence in metabolic profile of the evolved hosts was largely associated with a decrease in intermediates of central metabolism. A decrease in the accumulation of intermediates can be potentially explained by two alternative mechanisms. First, a higher metabolic flux rate can increase pathway completion, which means that the abundance of pathway intermediates decreases but the production of end-products will increase (Ferea et al., 1999; Maharjan et al., 2007; Pfeiffer et al., 2001). Alternatively, a reduction of overall metabolic rate would reduce the abundance of both pathway end-products and intermediates (Ibarra et al., 2002; Lewis et al., 2010). Both of these can be indicative of increased metabolic efficiency if performance is not compromised (Ratcliffe and Shachar‐Hill, 2006; Rees and Hill, 1994), which is the case here. In addition, the shared metabolic profile of the evolved hosts had decreased in antioxidant production. A reduction of antioxidants has been documented as an adaptation to stable photosymbioses in a number of cases, including in the Hydra - Chlorella symbioses (Ishikawa et al., 2016), and a similar decrease in host oxidative stress responses was observed through transcriptome comparisons of symbiotic compared to symbiotic-free P. bursaria (Kodama et al., 2014). This reduction in antioxidants is indicative of less oxidative stress within the hosts and is believed to be because symbionts take over the oxidative stress response resulting in a more tightly integrated symbiosis (Hörtnagl and Sommaruga, 2007; Summerer et al., 2009). In line with this, the reduction in host antioxidants was accompanied by an increase in symbiont oxidative stress protection in most of the symbiont replicates, specifically a photo-protective carotenoid and secondary metabolites with potential antioxidant properties.

In addition, in the evolved host metabolism of some replicates, I observed increased levels of algae-cell degradation components that could be indicative of increased Chlorella digestion. Symbiont digestion is an important element of symbiont load control that also provides nutrition for the host (Kodama and Fujishima, 2008; Titlyanov et al., 1996). Host digestion in response to a changing benefit-to-cost ratio has been documented in the cereal weevil whose digestion of its Sodalis endosymbiont changes with its benefit, and therefore alters throughout the developmental stages (Vigneron et al., 2014); and in corals that digest their dinoflagellate symbionts when starved (Titlyanov et al., 1996). Increased

108 digestion was observed in host replicates with either symbiont genotype and, therefore, was not specific to either a decrease or increase in symbiont load. Instead, symbiont digestion appears to be a general mechanism by which hosts can derive additional benefit from their symbionts.

Partner switching is a crucial aspect of endosymbioses that is known to rescue endosymbioses where the symbiont has lost key functionality or is mis-matched to new environmental conditions (Joy, 2013; Koga and Moran, 2014; Lefèvre et al., 2004; Matsuura et al., 2018). Due to a lack of prior coadaptation, successful partner switching may often require for an initial period of low fitness to be overcome. In this chapter I have used a powerful combination of physiological, metabolic and evolutionary methodologies to study the processes that underlie symbiosis integration. With this approach, I have demonstrated that a novel, initially non-beneficial symbiosis rapidly evolved to be beneficial, primarily through adaptations in host metabolism and symbiont load regulation. The host adapted to the novel symbionts by converging metabolically to the hosts with the native symbiont, but the symbiont load regulation between the hosts remained different, possibly connected to the alternative metabolic profile in the majority of the novel symbionts. Interestingly, the fitness of the novel pairings increased in all of the replicates despite the potentially different degrees of symbiont benefit. This could be because there are two alternative strategies regarding symbiont metabolism and symbiont load, and both strategies lead to higher fitness. Alternatively, it could reveal asymmetry between the contribution of the partners to adaptation, and that host adaptation is the primary driver in this endosymbiosis. Asymmetry between the contribution of host and symbiont to adaptation has been documented within other endosymbioses (Koch et al., 2017), and is theorised to be an aspect of their unequal control (Frank, 1997). The control within this endosymbiosis is thought to be especially one-sided since this photosymbiosis is believed to be an instance of host exploitation (Decelle, 2013; Lowe et al., 2016). Overall, these results support the hypothesis that rapid evolution of benefit can stabilise novel associations and so enables partner switching to occur with a broader range of partners than initial compatibility tests would reveal.

109

4.5 Supplementary Tables Table S4.1. Identified metabolites associated with PCA trajectories for the P. bursaria fraction. These were identified from the top 1% of loadings when using the first three principal components. The metabolite ID is that referred to in Figure 4.6. Detected Accurate Kegg / PC of loading ID mass mass Adduct Function Pathway Compound Metacyc PC1, PC3 X110.14 110 109.0197 H+ Amino acid Taurine metab Hypotaurine C00519 PC1 X170.407 170 131.0582 K+ Amino acid + Heme Heme biosynthesis 5-Amino-4-oxopentanoate C00430 147.0532 Na+ Amino acid/Central Glutamate C00025 131.0582 K+ Amino acid Glutamate 5-semialdehyde C01165 PC1, PC2, PC3 X265.213 265 226.0477 K+ Amino acid Shikimate pathway Chorismate C00251 226.0477 K+ Shikimate pathway Prephenate C00254 242.0192 Na+ Shikimate pathway Deoxy-ketofructose-phosphate C16848 PC1, PC2, PC3 X376.2.241 376.2 353.1699 Na+ Plant hormone Plant hormone (zeatin) Dihydrozeatin riboside C16447 PC1, PC2, PC3 X628.3.437 628.3 627.2487 H+ Plant degradation Chitin degradation Chitotriose CPD-13227 PC1, PC2 X629.3.437 629.3 628.2897 H+ Plant degradation Chlorophyll degradation Chlorophyll catabolite C18098 PC1, PC2 X664.3.420 664.3 625.3033 K+ Lipid Lipid - Arachidonic acid Leukotriene C4 C02166 PC1, PC2, PC3 X685.3.422 685.3 684.3178 H+ Antibiotic Antibiotic gamma-L-Glutamyl-butirosin B C18005 PC1, PC2 X686.3.421 686.3 685.3256 H+ Antibiotic Antibiotic Viomycin C01540 PC1 X737.4.422 737.4 714.3979 Na+ Antibiotic Antibiotic Avermectin B1b monosaccharide C11965 PC1, PC3 X803.4.369 803.4 780.3622 Na+ Antioxidant Glutathione metabolite Bis(glutathionyl)spermine C16563 PC2 X122.403 122 121.0197 H+ Amino acid Amino acid Cysteine C00736 PC3 X127.124 127 88.0160 K+ Central TCA/Glycolysis Pyruvate C00022 104.0110 Na+ Amino acid Hydroxypyruvate C00168 PC3 X171.66 171 132.0059 K+ Central Central/TCA/Glycolysis Oxaloacetate C00036 169.9980 H+ Glycolysis/Carbohydrate Glycerone phosphate C00111 169.9980 H+ Glycolysis/Carbohydrate Glyceraldehyde 3-phosphate C00118 132.0535 K+ Amino acid L-Asparagine C00152 PC3 X185.112 185 146.0215 K+ Central Central/TCA/amino acids 2-Oxoglutarate C00026 146.0579 K+ Pantothenate + CoA 2-Dehydropantoate C00966 146.0579 K+ Amino acid 2-Aceto-2-hydroxybutanoate C06006 PC3 X205.124 205 182.0579 Na+ Antioxidant Amino acid/antioxidant 4-Hydroxyphenyllactate C03672 182.0215 Na+ Antibiotic 3;5-Dihydroxyphenylglyoxylate C12325 166.0491 K+ Purine alkaloid Methylxanthine C16353

110

Table S4.1 continued Detected Accurate Kegg / PC of loading ID mass mass Adduct Function Pathway Compound Metacyc PC3 X220.282 220 181.0739 K+ Amino acid Amino acid Tyrosine C00082 181.0739 K+ Amino acid N-Hydroxy-L-phenylalanine C19712 PC3 X289.1.279 289.1 288.0998 H+ Antibiotic Antibiotic 6-Deoxydihydrokalafungin C12435 PC3 X377.2.241 377.2 354.2406 Na+ Lipid Lipid - Arachidonic acid Amoglandin C00639 PC3 X393.2.241 393.2 354.2406 K+ Lipid Lipid - Arachidonic acid Amoglandin C00639 370.2355 Na+ Lipid - Arachidonic acid 6-Keto-prostaglandin F1alpha C05961 370.2355 Na+ Lipid - Arachidonic acid Thromboxane B2 C05963 PC3 X398.1.241 398.1 359.1151 K+ Antibiotic Antibiotic Penicillin N C06564 397.0798 K+ Antibiotic 4-Ketoanhydrotetracycline C06627

111

Table S4.2. Identified metabolites associated with PCA trajectories for the Chlorella fraction. These were identified from the top 1% of loadings when using the first three principal components. The metabolite ID is that referred to in Figure 4.7.

Detected Accurate Kegg / PC of loading ID mass mass Adduct Function Pathway Compound Metacyc PC1, PC2, PC3 X122.403 122 121.0197 H+ Amino acid Amino acid Cysteine C00097 PC1, PC3 X393.2.370 393.2 354.2406 K+ Lipid Arachidonic acid Amoglandin C00639 370.2355 Na+ Arachidonic acid 6-Keto-PGF1a C05961 370.2355 Na+ Arachidonic acid Thromboxane B2 C05963 PC1, PC2 X421.2.241 421.2 382.2508 K+ Carotenoid + ABA Carotenoid + ABA synthesis C25-Allenic-apo-aldehyde C14044 PC1, PC3 X628.3.438 628.3 627.2487 H+ Cell wall metab Chitin degradation Chitotriose CPD-13227 PC1, PC3 X629.3.437 629.3 628.2897 H+ Chlorophyll degradation Chlorophyll degradation Chlorophyll catabolite C18098 PC1, PC2, PC3 X664.3.422 664.3 625.3033 K+ Lipid Arachidonic acid Leukotriene C4 C02166 PC2, PC3 X376.2.241 376.2 353.1699 Na+ Plant hormone Plant hormone (Zeatin) Dihydrozeatin riboside C16447 PC2 X607.3.420 607.3 568.305 K+ Chlorophyll Chlorophyll metabolism Protoporphyrinogen IX C01079 584.2635 Na+ Chlorophyll metabolism Bilirubin C00486 PC3 X217.35 217 178.0477 K+ Secondary metabolite Ascorbate/Vitamin C L-Galactono-1;4-lactone C01115 194.0579 Na+ Phenylpropanoid/cell walls Ferulate C01494 216.0399 H+ Isoprenoid biosynthesis 2-C-Methyl-D-erythritol 4-phosphate C11434 178.063 K+ Phenylpropanoid/cell wall Coniferaldehyde C02666 PC3 X299.432 299 260.0297 K+ Monosaccharide Starch + sucrose Glucose 6-phosphate C00092 260.0297 K+ phosphate Glycolysis Glucose 1-phosphate C00103 260.0297 K+ Fructose and mannose Mannose 6-phosphate C00275 276.0246 Na+ Pentose phosphate pathway 6-Phospho-D-gluconate C00345 260.0297 K+ Galactose Galactose 1-phosphate C00446 260.0297 K+ Fructose and mannose Mannose 1-phosphate C00636

112

Table S4.3. Change in symbiont load for each HK1 replicate between the start and end of the evolution experiment. The metabolic group column denotes whether the replicate’s metabolic profile converged with the profile of the native 186b symbionts or diverged. From these two groups (‘converge’ or ‘diverge’) a group mean difference in symbiont load was calculated.

Difference in Group mean HK1 replicate Metabolic group symbiont load difference 1 145404.2 converge

2 337137.9 converge 241271 3 745804.2 diverge

4 426775.4 diverge

5 490066.7 diverge 500951.4 6 341159.3 diverge

113

Chapter 5

Discussion

Endosymbiosis is an important evolutionary process underpinning a major evolutionary transition that has had a profound effect on the evolution of complex life (Keeling, 2010; Martin et al., 2015) and continues to impact the functioning of modern ecosystems (Baker, 2003; Powell and Rillig, 2018; Zook, 2002). The merger of two organisms can drive biological innovation (Wernegreen, 2012) and fundamental shifts in nutritional strategy, such as the transition to mixotrophy in photosymbioses (Esteban et al., 2010). Despite their importance our knowledge of these relationships remains limited, although there is increasing interest in this research area (Raina et al., 2018). In this thesis I have used experiments with the tractable microbial photosymbiosis between P. bursaria and Chlorella to study the mechanisms of host-symbiont specificity and partner switching.

First, I used a novel metabolic approach to compare the metabolic mechanism of two independent origins of the P. bursaria - Chlorella endosymbiosis. I found that convergence had occurred for the primary nutrient exchange, but that the metabolic mechanisms of light management had diverged and that these differences led to phenotypic variation. Next, I investigated the genetic variation in greater detail using a reciprocal cross-infection experiment coupled with metabolomics. I found that the differences in light-dependent symbiont stress responses affected the outcome of the interaction between host and symbiont genotypes, and thus underlies host-symbiont specificity. Finally, using an evolution experiment I found that host-symbiont specificity could be overcome as a novel association, initially lacking benefit, evolved to become a beneficial symbiosis. This data suggest that newly formed host-symbiont pairings can rapidly evolve higher fitness through changes in symbiont load regulation and metabolism. The capability for rapid partner amelioration and integration demonstrates the potential scope and flexibility for partner identity within partner switching.

In this general thesis discussion, I will explore some of the key themes emerging from my results, discuss potential applications of endosymbiosis research, and suggest future directions that could expand on my findings.

114

5.1 Stress and symbiosis Light is a key factor mediating the fitness effects of photosymbiosis, and its dual role as both the source of energy and of potentially damaging agent of oxidative stress is well documented (Decelle et al., 2015; Venn et al., 2008; Yakovleva et al., 2009). Across this thesis, my findings have shown how critical light management is for the P. bursaria - Chlorella endosymbiosis, and that variation in light management affects both the fitness and compatibility of host-symbiont pairings. In Chapter 2, I found that the Chlorella strains from the two independent originations of the symbiosis had diverged in their light stress tolerance; the HA1 genotype increased production of photoprotective compounds in response to high light, while the 186b genotype instead invested more in photosynthetic machinery enabling high irradiance levels to be used effectively in photosynthesis. Further differences in light-associated stress responses among the clades were discovered in Chapter 3. Whereas the 186b Chlorella displayed a dark-associated stress-response, the HA1 and HK1 Chlorella displayed high-light-associated stress responses. These responses translated into higher costs of symbiosis in the dark for the 186b host-symbiont pairing, but lower benefits of symbiosis in high light for the HA1 host-symbiont pairing. Together these results suggest a key role of light-associated stress tolerance in the context dependent fitness effects of symbiosis and as a cause of divergence among the independent originations of this symbiosis. This aligns with other photosymbioses, particularly studies in the coral - Symbiodinium and Hydra - Chlorella endosymbioses, where thermal and light stress tolerance determine the fitness outcome of the symbiosis (Abrego et al., 2008; Ye et al., 2019).

Light stress tolerance is a product of the host-symbiont interaction. For example, in Chapter 3 I show that the dark-associated stress response induced in the 186b Chlorella in their native background is alleviated when 186b Chlorella are residing in the HA1 host genotype background. This suggests a hitherto unknown mechanism of host-mediated amelioration that could possibly be linked to greater provisioning preventing symbiont starvation. In addition, in Chapter 4 I show that an initially non-beneficial novel host- symbiont association could rapidly evolve to become beneficial in part due to changes in expression of stress-related metabolites in both the host and the symbiont. These results have implications for the hypothesis of Kawano et al. (2004) who theorised that P. bursaria, in contrast to other Paramecium species, acquired symbionts because of their pre- adaptations to oxidative stress. The results of this thesis support this idea but show that different P. bursaria genotypes vary in their ability to tolerate and ameliorate stress with consequences for host-symbiont specificity. In particular, it is notable that host genotype

115 backgrounds, such as HA1, that appear capable of alleviating symbiont stress are better able to establish beneficial symbioses with a wider range of symbiont genotypes, i.e., they are generalists in terms of specificity.

5.2 Partner Switching Partner switching can rescue symbioses by restoring symbiont function (Koga and Moran, 2014; Matsuura et al., 2018), enable rapid adaptation to environmental change (Boulotte et al., 2016; Lefèvre et al., 2004), and facilitate niche-expansion (Joy, 2013; Rolshausen et al., 2018; Sudakaran et al., 2017). In particular, local adaptation by symbiont acquisition is likely to occur far faster than by symbiont evolution and may therefore be an important mode of host adaptation. In the majority of systems, however, the mechanisms that enable and restrict partner-switching have rarely been elucidated.

The Chlorella clades associated with the European and Japanese/American originations of the P. bursaria-Chlorella symbiosis are highly diverged, indeed, the European Chlorella clade is more closely related to the Hydra-symbiotic Chlorella than the Chlorella from the Japanese/American clade of the P. bursaria-Chlorella symbiosis (Hoshina et al., 2005). Nevertheless, while algal symbiont switching between Hydra and European clade P. bursaria is not possible (Summerer et al., 2007), I show that partner-switching between the European and Japanese/American clades of the P. bursaria - Chlorella symbiosis results in functional host-symbiont pairings. This partner-switching is enabled by a convergent nutrient exchange between these two originations of the P. bursaria - Chlorella endosymbiosis (Chapter 2). Both Chlorella from Hydra and P. bursaria are thought to supply their hosts with maltose (Mews, 1980; Ziesenisz et al., 1981), but it appears likely that the nitrogen source they receive in return differs. In Hydra, Chlorella appear to receive glutamine from their host (Hamada et al., 2018), while in P. bursaria my results in Chapter 2 suggest that the nitrogen source is arginine. This divergence in N-exchange metabolite could drive the observed between-species incompatibility (Summerer et al., 2007). Many endosymbioses have multiple independent origins, and convergence upon shared symbiotic exchanges despite genetic differences appears to be a common theme (Gargas et al., 1995; Masson-Boivin et al., 2009; Sandström et al., 2001). This suggests that metabolic function, and not simple genetic identity, may underlie successful partner-switching, which in turn will determine which novel combinations of host and symbiont can establish new associations.

116

I have shown that the P. bursaria host genotypes vary in their degree of partner-generalism, seen by the symbiotic fitness of novel associations (Chapters 2 & 3). Specifically, my findings show that genotypes varied such that one host genotype appeared to be a generalist (HA1), another a specialist (186b), and the third was intermediate (HK1) for partner specificity. The influence of host-symbiont genotype interactions on the outcome and fitness effects of symbiosis is a core component of co-evolutionary theory (Thompson, 2005). Coexistence of generalist and specialist strategies for partner specificity suggests that each strategy may confer fitness benefits (Wilson and Yoshimura, 1994). This is typically explained by the “Jack-of-all trades is a master of none” hypothesis, wherein although generalists occupy a wider range of niches, specialists have higher fitness in their chosen niche (Futuyma and Moreno, 1988; Straub et al., 2011). Generalist-specialist coexistence is seen across diverse symbioses. For example variation in the degree of host generalism has been reported in legume and rhizobia associations (Wilkinson and Parker, 1996), ectomycorrhiza and conifer symbioses (Molina and Trappe, 1982), and Symbiodinium-hosting corals, with absolute partner specificity being rarely observed (Silverstein et al., 2012). Indeed, some host species possess a remarkable degree of partner- generalism, for instance western hemlock forms associations with over 100 fungal symbiont species (Kropp and Trappe, 1982). In contrast to the “Jack-of-all trades” hypothesis, the P. bursaria – Chlorella specialist host examined here did not have higher fitness than the generalist host when tested with its conspecific symbiont. This implies that this specialist strain does not represent an alternative evolutionary optimum, but rather would be outcompeted if the strains co-occurred naturally. Furthermore, generalist host genotypes are more likely to successfully integrate new symbiotic partners. However, interestingly host generalism was associated with a consistent symbiotic phenotype across diverse algal symbionts, suggesting that such hosts may be less able shift their ecological niche through partner switching. On the other hand, specialism can preclude adaptation and absolute dependency on partner genotype can prevent partner switching entirely (Moran and Wernegreen, 2000). My results provide evidence that host genotype has a large influence determining the compatibility of new symbiotic partners and therefore the potential for partner switching.

This thesis has increased the understanding of the metabolic mechanisms underlying partner switching, which are important to understand from both an evolutionary and ecological perspective. For the former, a greater understanding of partner switching is required if we are to understand life-history patterns and establish an accurate understanding of the eukaryotic tree of life (Keeling, 2010). Specifically, the spread of

117 plastids has involved serial symbiont replacement that has entangled lineages (Delwiche, 1999; Dorrell and Smith, 2011; Stiller et al., 2014). The ‘shopping bag model’ hypothesises that the replaced symbiont can have transferred genes to the host nucleus, leading to a complement of endosymbiont genes and proteins from mixed origins (Larkum et al., 2007; Patron et al., 2006). These replacements have involved significant transitions between red and green plastids, but we do not currently know the number of these replacements and whether they were functionally equivalent or endowed novel traits (Archibald and Keeling, 2002). For ecological systems, partner-switching will affect ecosystems by potentially enabling migration and adaptation to environmental change, and because endosymbioses are often keystone organisms, changes in these relationships will have knock-on effects (Zook, 2002). For instance, partner-switching has been an important factor in insect endosymbioses diversification (Sudakaran et al., 2017) and in some instances symbiont replacement has been associated with nutritional transitions (Bell-Roberts et al., 2019) that have had large effects on the plants the insects feed on and their many predators and competitors (Frago et al., 2012; Sugio et al., 2015).

5.3 Rapid evolution enables the establishment of symbiosis A key finding of this thesis is that initially non-beneficial novel symbiotic pairings can rapidly evolve to become beneficial (Chapter 4). This suggests that following partner- switching a period of adaptation may often be required to allow for the novel symbiont to be integrated. There are multiple examples that have documented the rapid evolution of the fitness outcome of a symbiosis, though most of the previous examples have focused on transitions between parasitism and mutualism (King et al., 2016; Sachs and Wilcox, 2006; Shapiro and Turner, 2018; Tso et al., 2018). For instance, the opportunistic fungal pathogen Candida albicans was found to quickly evolve to protect its mouse host from systemic infections (Tso et al., 2018), demonstrating that initially costly symbionts could evolve to be beneficial. In my results the evolution of benefit appeared to be predominately driven by the evolution of host metabolism and symbiont-load regulation, suggesting asymmetry in the contribution of the partners to this adaptation. The relative contribution of host versus symbiont adaptation to the evolution of a symbiosis varies (Hill, 2009; Koch et al., 2017), and is partially determined by the relative level of control each partner exerts (Frank, 1997; Johnstone and Bshary, 2002) as well as other factors such a relative generation time. In a parasitism, like the starting condition used by Tso and colleagues, the symbiont initiates the interaction and has more control. The symbiont is predicted, therefore, to have the dominant contribution to adaptation, and it was the evolution of the fungus that drove the transition from parasitism to mutualism in the example above

118

(Tso et al., 2018). In contrast, the P. bursaria - Chlorella symbiosis is an example of host exploitation (Decelle, 2013; Lowe et al., 2016) and therefore, the host initiates the interaction and has more control. This could explain why host adaptation was the primary driver in the evolution experiment of Chapter 4.

5.4 Applications of endosymbiosis research Endosymbioses have keystone roles in ecosystems, and as such knowledge of these associations can have implications for conservation and agriculture (Anthony et al., 2017; Monika et al., 2019). It is becoming increasingly urgent that we understand how symbiotic interactions respond to climate change, which alters the environmental context and therefore potentially the fitness outcome and stability of symbiotic associations (Kikuchi et al., 2016; Stat et al., 2006; Thompson, 2005). In response to rapid environmental change endosymbioses must either adapt to the new conditions, switch to partners better adapted to the new conditions, or otherwise they may face extinction. A particular focus has been on the breakdown of the coral-Symbiodinium endosymbiosis with increased ocean temperature that leads to coral bleaching (Douglas, 2003; Weis, 2008), which in turn can lead to coral mortality that has already caused catastrophic loss of coral reefs (Hughes et al., 2017; Sully et al., 2019). Investigations have found that the survival of this key endosymbiosis can be increased through the survival of partner-generalist hosts and increased levels of symbiont diversity (Fabina et al., 2013). Crucially, partner switching to more thermally tolerant symbionts can enable survival of the corals (Berkelmans and van Oppen, 2006; Rowan, 2004), and there are hopes that this will prevent the complete loss of these habitats (Coles et al., 2018; Hughes et al., 2003). Furthermore, there have been recent calls to use directed evolution to introduce more tolerant symbionts and induce beneficial partner switching in order to promote the survival of this keystone endosymbiosis (Anthony et al., 2017; van Oppen et al., 2015, 2017). My results have shown similar symbiont-genotype variation in stress tolerance within the P. bursaria - Chlorella symbiosis, and, importantly, show that benefit can rapidly evolve in novel associations. The implication of which is that testing novel coral - Symbiodinium associations should include sufficient time for these novel pairings to co-adapt, because an initial lack of co- adaptation may prevent new phenotypic properties being immediately evident.

Endosymbiotic research has also recently been applied to agriculture and health management. Specifically, the potential for using plant-microbe interactions to improve crop yield has received a lot of interest; both in terms of manipulating mycorrhizal fungi endosymbioses to act as biofertilizers (Monika et al., 2019) and for using a broader range

119 of rhizobacteria to increase plant protection against pathogens and pests (Kour et al., 2019). In addition, the disruption of key endosymbiotic associations of insects is being considered for a symbiosis-based method of pest control (Hosokawa et al., 2007; Nobre, 2019). Finally, endosymbiotic research is already being used in parasite control and Wolbachia infected mosquitos have been released into the wild in Brazil, Florida and Australia (O’Neill, 2018). The Wolbachia endosymbionts are used to inhibit the infection of human-disease causing viruses or parasites to stop the spread of mosquito-borne diseases such as dengue, Zika and malaria (Bourtzis et al., 2014; Caragata et al., 2016; Werren et al., 2008). My results have shown that metabolic function underlies the basis of partner compatibility, and this may be a useful consideration when designing artificial symbiotic partnerships for these applications.

The manipulation of endosymbioses for conservation, health or agriculture requires a detailed understanding of the mechanistic basis of these interactions in order that our alterations can have the desired effect. Understanding these systems requires the combination of controlled laboratory-based research to study the underlying mechanisms and large-scale in-situ studies that can examine the role of complex multi-species interactions within their ecosystems. The P. bursaria - Chlorella photosymbiosis is analogous in many ways to the coral - Symbiodinium endosymbiosis, and this thesis has demonstrated how stress response and genotype variation interactions can be understood at a mechanistic level in this simpler, microbial endosymbiosis. Future work could develop the P. bursaria - Chlorella association as a model for the less tractable coral - Symbiodinium relationship and use the tractable microbial system to explicitly test hypotheses that cannot currently be tested in the more complex system.

5.5 Future-directions Following on from the work in this thesis on the mechanisms that underlie host-symbiont specificity, I believe the next logical step would be to directly test partner switching within this system. Re-infection experiments with a diverse group of symbiont genotypes could determine whether P. bursaria displays active partner choice, and if so whether the conspecific symbiont is chosen or if the most beneficial symbiont for the local conditions is selected.

Future work could also use a greater variety of genotypes to establish whether the mechanistic patterns observed here are representative. Specifically, it would be necessary to include other European clade strains to separate clade versus strain effects. This data

120 could test whether generalist and specialist strategies for partner specificity co-exist within the main biogeographical clades. In addition, strain natural habitat data could be incorporated and tested for correlations with symbiosis specificity or light ecotype traits.

The potential presence of bacterial endosymbionts within P. bursaria was not considered within this thesis, though it is highly likely that they are present (Fokin, 2004; Gong et al., 2014). Future work could build on the current candidate bacterial symbionts and identify which species are stably present within these systems and whether the bacterial community differs between symbiont-free and symbiotic P. bursaria. It may be that we need to view the P. bursaria - Chlorella relationship as one component of a complex multi- partner consortium.

Furthermore, greater molecular knowledge is required, particularly Paramecium specific metabolomic, transcriptome and genomic data should be increased and integrated to produce curated databases. Further integration of genomic data is needed for both partners, and it would be interesting to compare full genome sequences of the two symbiont clades to uncover the genetic differences that underpin their metabolic variation. Moreover, increased molecular knowledge of this system would allow the development of genetically transformed partners that would allow specific hypotheses to be tested.

5.6 In conclusion Independent origins of symbiosis is common (Masson-Boivin et al., 2009; Muggia et al., 2011; Sandström et al., 2001) and I have found that these provide important standing variation that enables phenotypic diversity, but that metabolic compatibility, not genetic identity, defines the limits of partner integration. This suggests that efforts to study the diversity of symbiotic interactions need to include metabolic function. Although the use of metabolomics within symbiosis is growing (Achlatis et al., 2018; Chavez-Dozal and Nishiguchi, 2016; Padfield et al., 2016), there is an over-reliance on isolated sequence studies. This holds particularly true for microbiome research’s tendency to use 18S and 16S rRNA sequencing results without integration of metabolites, transcripts and proteins (Knight et al., 2019; Poretsky et al., 2014).

Stress is known to be a critical factor for partner integration within photosymbioses (Howells et al., 2012; Venn et al., 2008; Ye et al., 2019). My results add to this knowledge by showing that light-dependent symbiont stress-responses drive host-symbiont genetic specificity within the P. bursaria - Chlorella association. This reveals that this microbial

121 symbiosis could be used as a model photosymbiosis to investigate the role of stress further, which is of particular relevance to the coral - Symbiodinium association. In addition, this thesis explicitly links stress tolerance and partner specificity, suggesting that host genotypes best able to alleviate stress in their symbionts are the most generalist in terms of the fitness of their associations with non-native symbiont genotypes. Typically, these two factors are considered separately and a potential link between the two needs to be investigated to test if it is common within photosymbioses. If true, this would imply stress tolerance is one of the critical factors in the establishment of these relationships and would build on Kawano et al.’s (2004) hypothesis that it is a necessary pre-adaptation. This would then become an important consideration in the transition to endosymbiosis and might have implications for our understanding of plastid acquisition (Lesser, 2006).

This thesis has shown that a novel association can rapidly evolve to be beneficial and so overcome an initial lack of co-adaptation and benefit common to novel associations (Matthews et al., 2018; McGraw et al., 2002; Russell and Moran, 2005). This is an important finding for understanding the dynamics of partner switching and reveals that partner integration can occur in a broader range of circumstances than initial compatibility tests would reveal. This has ramifications in the growing field of endosymbiosis manipulation (Anthony et al., 2017; Monika et al., 2019), because it shows that new phenotypes may only become visible once a novel association has had time to co-adapt. This is particularly salient for the current efforts to rescue the coral - Symbiodinium symbiosis through symbiont replacement to more thermally-tolerant symbiont genotypes (van Oppen et al., 2015, 2017).

Endosymbiosis requires experimental research to fully understand the diversity and complexity of these intimate symbiotic interactions. This work highlights the power of metabolomics to characterise the mechanistic basis of partner specificity and partner switching. Endosymbiotic relationships involve the integration of two organism at every level of their biology, from their ecology, metabolism, genetics and evolution, and if we are to understand these complex interactions in their entirety, we must integrate our levels of study as we move forward.

122

Appendix A – The review paper from which extracts were taken for the Introduction (Chapter 1)

123

FEMS Microbiology Letters, 366, 2019, fnz148

doi: 10.1093/femsle/fnz148 Advance Access Publication Date: 4 July 2019 Minireview Downloaded from https://academic.oup.com/femsle/article-abstract/366/12/fnz148/5528313 by University of Sheffield user on 09 August 2019

M I N I REV I EW – Incubator The role of exploitation in the establishment of mutualistic microbial symbioses Megan E. S. Sørensen1,*,†, Chris D. Lowe2,EwanJ.A.Minter1,A. Jamie Wood3,4, Duncan D. Cameron1 and Michael A. Brockhurst1,‡

1Department of Animal and Plant Sciences, University of Sheffield, Sheffield S10 2TN, UK, 2Centre for Ecology and Conservation, University of Exeter, Penryn Campus, Cornwall TR10 9FE, UK, 3Department of Biology, University of York, York YO10 5DD, UK and 4Department of Mathematics, University of York, York YO10 5DD, UK

∗Corresponding author: Department of Animal and Plant Sciences, University of Sheffield, Sheffield, S10 2TN, UK. Tel: +0114 222 0140; E-mail: [email protected] One sentence summary: The authors review the theoretical and experimental evidence supporting exploitation as an alternative route to the evolution of mutualistic symbioses. Editor: Daniel Tamarit †Megan E. S. Sørensen, http://orcid.org/0000-0001-8983-2943 ‡Michael A. Brockhurst, http://orcid.org/0000-0003-0362-820X

ABSTRACT Evolutionary theory suggests that the conditions required for the establishment of mutualistic symbioses through mutualism alone are highly restrictive, often requiring the evolution of complex stabilising mechanisms. Exploitation, whereby initially the host benefits at the expense of its symbiotic partner and mutual benefits evolve subsequently through trade-offs, offers an arguably simpler route to the establishment of mutualistic symbiosis. In this review, we discuss the theoretical and experimental evidence supporting a role for host exploitation in the establishment and evolution of mutualistic microbial symbioses, including data from both extant and experimentally evolved symbioses. We conclude that exploitation rather than mutualism may often explain the origin of mutualistic microbial symbioses.

Keywords: microbiology; experimental evolution; microbial symbioses

INTRODUCTION of cheating are expected to outweigh the potential long-term fitness benefits of cooperation, producing a ‘tragedy of the com- Symbiosis – ‘the living together of unlike organisms’(De Bary mons’ (Hardin 1968; Rankin, Bargum and Kokko 2007). Therefore, 1879) – encompasses a broad range of species interactions, both in long-established associations and in the establishment including both parasitism (+/– fitness interactions) and mutu- of new relationships, evolutionary conflict and breakdown of alism (+/+ fitness interactions). Whilst the evolutionary ratio- mutualistic symbiosis is ever likely, since each partner is under nale for parasitism is straightforwardly explained by the self- selection to minimise its investment in the integrated symbi- interest of the parasitic partner, explaining the origin of mutual- otic unit (Perez and Weis 2006; Sachs and Simms 2006). Never- istic symbiosis is more challenging. The immediate fitness gains theless, mutualistic symbiotic relationships are abundant, taxonomically widespread, ecologically important in a wide range of

Received: 21 February 2019; Accepted: 1 July 2019

C FEMS 2019. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.

1 2 FEMS Microbiology Letters, 2019, Vol. 366, No. 12

habitats, economically important in agricultural systems and, otherwise grow faster outside of symbiosis. This is a special case consequently, underpin the biodiversity and function of both of parasitism known as host exploitation, which has been far natural and man-made ecosystems (Bronstein 2015;Powelland less well-studied. In this review, we gather together the evidence Rillig 2018). supporting a role for host exploitation in the establishment of Mutualistic symbiosis can accelerate evolutionary innova- mutualistic microbial symbiosis. tion through the merger of once independent lineages, providing species with new ecological traits and allowing them to inhabit previously inaccessible ecological niches (Wernegreen THEORETICAL STUDIES OF SYMBIOSIS: Downloaded from https://academic.oup.com/femsle/article-abstract/366/12/fnz148/5528313 by University of Sheffield user on 09 August 2019 2004; Kiers and West 2015). A classic example of this is nutrient MUTUALISM VERSUS EXPLOITATION trading, where the partners exchange compounds that are oth- The paradox of mutualism erwise difficult or impossible for them to acquire. These include aphids with their obligate endosymbiont Buchnera aphidicola Mutualisms are abundant throughout the tree of life despite that exchange essential amino acids (Moran et al. 2003), and their inherent evolutionary conflicts, and this disparity is con- land plants with arbuscular mycorrhizal fungi where fixed car- sidered the paradox of mutualism. The paradox of mutualism bon is exchanged for phosphate and organic nitrogen (Pfef- has been well explored using theoretical models that aim to fer et al. 1999). Besides exchanging nutrients, mutualistic sym- discover the evolutionary stable strategies of mutualistic symbioses can involve a wide range of benefits, including the pro- biosis. The reciprocal exchange of services/goods within mutu- duction of antibiotics (Currie et al. 1999), luminescence (Tebo, alisms make them a specific form of group cooperation. There Scott Linthicum and Nealson 1979), photoprotection (Hortnagl¨ are two primary evolutionary explanations for group coopera- and Sommaruga 2007) and protection from predation (Tsuchida tion. Within a species, kin selection explains that helping related et al. 2010). Since many of these potential benefits may only be individuals provides inclusive fitness benefits to the actor (fol- required in particular environments or at particular times, many lowing Hamilton’s rule (Hamilton 1964)). Alternately for non- symbioses vary ecologically across a continuum from mutual- relatives, game theory has provided the strategic alliance model, ism to parasitism (Heath and Tiffin 2007; Wendling, Fabritzek which is based around reciprocity and includes the Tit-for-Tat and Wegner 2017). Indeed, some organisms may only engage in strategy (Axelrod 1984). Frank (1996), however, highlighted that symbiosis when in nutrient-deficient environments (Muscatine the evolution of interspecific symbiosis cannot be explained by and Porter 1977; Johnson 2011). either of these models; kin selection is not applicable because Mutualistic symbiosis involves a shift in individuality as the interaction is between unrelated individuals from different two unrelated species evolve inter-dependence and transition species, and the strategic alliance model fails because it requires to function as a single organism (Szathmary´ and Smith 1995; memory of past interactions, the recognition of individuals and Estrela, Kerr and Morris 2016). In nature, the degree of depen- is dissipated by forms of mixing. The traditional explanations dence varies extensively both within and between symbioses for cooperation are, therefore, insufficient to explain the evolu- (Minter et al. 2018). Dependence can range from obligate asso- tionary stability of symbioses. ciations with mutually dependent partners, through asymmet- Theoretical work has consequently focused on mutualism- rically dependent associations where only one species is unable specific explanations, and a key process underlying much of to survive alone, to fully facultative associations where both this work is finding mechanisms that align the partners’ fit- species can survive alone. Comparative studies suggest that ness interests. Herre et al. (1999) proposed that this alignment mutual dependence is more likely to evolve in vertically inher- could be achieved by ‘conflict avoidance factors’, which include ited symbioses, where the fitness interests of both species vertical transmission, genetic uniformity of symbionts, popula- become aligned (Fisher et al., 2017). Transitions in individuality tion spatial structure and obstructions to alternative free-living are, however, fraught with evolutionary conflict, and the merger states. The influence of these factors has been explored by the- of two independent organisms is rarely seamless and never self- oretical models, particularly vertical transmission that aligns less. Conflict is likely to be greatest during the establishment the reproductive interests of the partners (Yamaura (1993)). For of new symbioses, before the partners have been able to evolve reproductive interests to be fully aligned, both absolute co- complex mechanisms required to align their fitness interests. dispersal and reproductive synchrony are required as part of ver- Explaining the establishment of mutualistic symbioses is tical transmission (Frank 1997). If achieved, this reduces within- therefore challenging, and this is the focus of our review. As host competition between symbionts and stabilises the mutu- we shall explain in the subsequent section, the conditions for alism because the reproductive success of the symbiont is per- mutualistic symbioses to establish through mutualism alone fectly aligned to that of its host. Vertical inheritance is com- are highly restrictive, and thus several alternative mechanisms mon in well-established, obligate symbiotic partnerships and have been proposed (Garcia and Gerardo 2014; Keeling and is associated with greater dependence (Fisher et al. 2017). It is McCutcheon 2017). One of these is that mutualistic symbioses not, however, ubiquitous and there are many stable mutualisms evolve from parasitisms. This transition can occur in two direc- that maintain horizontal transmission. For example, Vibrio fis- tions. First, the smaller parasitic partner living in or on the larger cheri and bobtail squids (Visick and Ruby 2006), Rhizobia and host can evolve reduced virulence to eventually become benefi- legumes (Sprent, Sutherland and Faria 1987), and Endoriftia perse- cial to its host (King et al. 2016; Shapiro and Turner 2018; Tso phone and tube worms (Nussbaumer, Fisher and Bright 2006). et al. 2018). Sach et al. (2011) used phylogenetic reconstruction Consequently, it is clear that while conflict avoidance factors to predict whether bacterial symbionts originated as mutualists help to promote stability of some interactions, they are neither or parasites. For 42 beneficial bacterial symbionts, they inferred necessary nor sufficient for the evolutionary stability of mutual- that 32 had originated as parasitic whilst only 9 had originated istic symbioses (Genkai-Kato and Yamamura 1999). as mutualists (with 1 case remaining ambiguous), suggesting Frank (1995) provided a solution to the paradox of mutu- that parasitism is a more common route than mutualism to alism by developing a model centred on policing strategies, mutualistic symbiosis. Second, the larger host partner could which repressed competition and reduced the benefits of cheat- capture and exploit the smaller beneficial partner, which would ing to ensure the fair distribution of resources. Furthermore, the Sørensen et al. 3

results of the extended policing model (Frank 1996) showed that to escape the host, and the victim can become trapped in the variation in individual resources altered the degree of invest- symbiotic state. It is important to note that this interaction has ment in policing, with well-supplied individuals doubling their now become a mutualistic symbiosis; the victim provisions the policing investment and poorly supplied individuals not invest- host to the host’s benefit, whilst the victim’s reproductive rate ing at all. The theoretical prediction for the role of policing in symbiosis now exceeds that which is achievable in free-living in maintaining mutualistic symbioses has been supported by environments containing the host. numerous occurrences in a wide-range of natural systems. For Because host exploitation does not require symmetric example, partner sanctions in the legume–rhizobium symbio- mutual benefits at the outset nor complex stabilising mecha- Downloaded from https://academic.oup.com/femsle/article-abstract/366/12/fnz148/5528313 by University of Sheffield user on 09 August 2019 sis (Kiers et al. 2003), partner choice in the yucca–yucca moth nisms to allow establishment, it offers a simpler explanation symbiosis (Bull and Rice 1991), partner fidelity in solitary wasp– for the emergence of mutualistic symbiosis. Once mutualistic Streptomyces symbiosis (Kaltenpoth et al. 2014) and screening in symbiosis is established, further stabilising mechanisms could the bobtail squid–Vibrio fischeri symbiosis (McFall-Ngai and Ruby evolve to prevent its breakdown. Thus mutualism-stabilising 1991; Archetti et al. 2011). mechanisms may often be a secondary phenomenon, arising to Following Frank’s first policing models, there has been exten- further enforce originally exploitative but now mutualistic sym- sive development of theory exploring the evolution of mutu- bioses. alism. The current consensus is that stabilising mechanisms, such as the various policing strategies, vertical transmission and other conflict avoidance factors, provide solutions to the para- EXPLOITATION IN ACTION dox of mutualism (for extensive reviews of the topic, see Sachs Empirical data on the establishment of mutualistic symbioses et al. (2004); Leigh (2010) and Archetti et al. (2011)). However, while are rare because studying this process experimentally is chal- it is clear that these complex adaptations play a crucial role in lenging. The extant mutualistic symbioses we observe in nature the maintenance of extant mutualistic symbioses, it is unlikely are the products of co-evolution and no longer in the establish- that they can explain the origin of new symbioses because here ment phase. Furthermore, for obligate mutualistic symbioses it there is little time for such complex stabilising mechanisms to may be impossible to separate the partners and therefore untan- evolve. The pre-existence of such traits, allowing for their co- gle the costs/benefits that each of the symbiotic partners derive. option for the purpose of stabilising symbiosis, may be a pre- Nonetheless, there are several mutualistic microbial symbioses requisite for the establishment of symbiosis. For instance, one that are amenable to experimental study, and two main experi- can imagine that partner-choice could evolve from pre-existing mental approaches. The first approach is to study extant faculta- feedback mechanisms and may even provide the selective envi- tive associations that remain experimentally tractable and allow ronment from which the symbiosis establishes (Frederickson the direct measurement of the relative costs and benefits of both 2013). However, given that complex stabilising mechanisms are the free-living and symbiotic states. The second approach is to not ubiquitous this seems unlikely to be a general explanation. experimentally evolve newly formed symbioses in the labora- Moreover, elaborate host–symbiont interactions, such as the tory to explore the environmental conditions that promote their bobtail squid–Vibrio fisheri multistage screening process, must establishment and stability (Hoang, Morran and Gerardo 2016). have evolved subsequent to establishment, even if the funda- We review the data from both approaches in the following sec- mental aspects were pre-adaptations. It is more parsimonious tion. therefore to assume that important limitations exist as to the conditions where mutualism can act as an establishment mechanism for mutualistic symbiosis. Experiments with extant facultative mutualistic microbial symbioses Exploitation as an alternative route to symbiosis One of the best studied facultative mutualistic microbial sym- An alternative route to the establishment of mutualistic symbioses is that between the single-celled ciliate host Paramecium biosis was proposed by Law and Dieckmann (1998). This model bursaria and its green alga symbiont, Chlorella. This classical pho- predicted that exploitative relationships wherein a host exploits tosymbiosis is founded upon the exchange of fixed carbon from a ‘victim’ species which it acquires by horizontal transmission the photosynthetic algae in return for organic nitrogen from the can evolve into stable mutualistic symbioses with vertical trans- host (Fig. 1). It has been estimated that the Chlorella endosym- mission simply through natural selection to increase individual bionts release 57% of their fixed carbon to the host (Johnson fitness. The key requirement for this outcome was that the free- 2011), primarily as maltose (Ziesenisz, Reisser and Wiessner living victim pays a cost to defend itself from being captured by 1981). The nitrogen source is not yet verified; current candi- the host. In this scenario, there is a trade-off for the victim, who dates include amino acids (Kato, Ueno and Imamura 2006;Kato either uses resources to defend itself or to provision the exploita- and Imamura 2008b), nucleic acid derivatives (Soldo, Godoy and tive host. Depending on the relative magnitude of these trade- Larin 1978; Shah and Syrett 1984) and ammonia (Albers, Reisser offs, it is possible that the victim has higher fitness in symbiosis. and Wiessner 1982). In this case, the evolution of vertical transmission is advanta- Crucially, while the symbionts are inherited vertically with geous to both partners as the victim has a higher reproductive tight cell cycle synchrony, the partners can be separated by son- rate in symbiosis than when free-living, where it must pay a high ication/chemical treatment (Kodama and Fujishima 2008, 2011, cost of defence. However, it remains the case that the victim’s 2012) allowing the costs and benefits of symbiosis versus free- optimal state would be to be free-living with no interaction with living to be directly compared. For hosts, the benefit of sym- the exploiter and thus paying neither of these costs. The model biosis increases with light intensity, such that while it is costly demonstrated that if the trade-off is sufficiently strong, the evo- to harbour symbiotic algae in the dark (i.e. symbiont-free hosts lution of stable symbiosis can be advantageous to both partners grow faster than symbiotic hosts), these costs are outweighed even in an exploitative relationship. Furthermore, once vertical at higher light intensity such that symbiosis is highly beneficial transmission has evolved it becomes much harder for the victim for hosts in high light. In contrast, symbiosis is never beneficial 4 FEMS Microbiology Letters, 2019, Vol. 366, No. 12 Downloaded from https://academic.oup.com/femsle/article-abstract/366/12/fnz148/5528313 by University of Sheffield user on 09 August 2019

Figure 1. Paramecium bursaria and Chlorella endosymbiosis. A. Z-stack of confocal sections of the chlorophyll autofluorescence of Chlorella endosymbionts within one Paramecium bursaria cell. With colour representing the intensity of fluorescence and therefore the position of the Chlorella in the Z-plane. B. Diagram of the relationship, showing the nutrient exchange with the transfer of maltose from the Chlorella in exchange for organic nitrogen (denoted as ‘N’ as the identity of this compound is currently unknown). Ma = macronucleus; Mi = micronucleus. for the alga: free-living algal growth rates increase monotoni- uptake (Kato and Imamura 2008a, 2008b). If the symbiont’s mal- cally with light intensity and at all light levels exceed those of tose is broken down to glucose by the host, then this control symbiotic algae. Moreover, hosts impose tight control on algal process would facilitate a reward system for more co-operative symbiont load (i.e. the number of algal symbionts per host cell) symbionts. The multiple control processes identified to date are which peaks at low light, and is reduced both in the dark and at all host-derived, supporting the idea that this symbiosis was high light intensity (Lowe et al. 2016). A mathematical model of founded upon exploitation. the symbiosis showed that hosts manipulate symbiont load in Phylogenetic analysis shows that symbiotic and free-living this way to maximise their return from nutrient trading, effec- Chlorella form polyphyletic groups (Hoshina and Imamura 2008; tively minimising their nitrogen cost for each molecule of car- Summerer, Sonntag and Sommaruga 2008), indicating multiple bon they gain from their algal symbionts (Dean et al. 2016). transitions to and from symbiosis. Moreover, diverse isolates Indeed, measurements of algal photosynthetic efficiency sug- of P. bursaria–Chlorella vary in their degree of dependence; from gested that algal symbionts were more nitrogen-starved than completely facultative associations to obligate mutual depen- their free-living counterparts (Lowe et al. 2016). Similar patterns dence, via asymmetric dependence where hosts depend on sym- of cost:benefit and host control were observed across a range of bionts but not vice versa (Minter et al. 2018). Taken together, geographically diverse isolates (Minter et al. 2018). these experimental data suggest that the nutrient trading rela- The mechanism of the control in this relationship is likely to tionship between the ciliate and the alga is exploitative rather be multifaceted, but in large part is thought to be due to host than mutualistic, benefiting the host (Lowe et al. 2016). Addi- digestion. Host selection in the establishment of the symbio- tional selective forces may be required therefore to explain the sis specifies which Chlorella are packaged into vacuoles and re- benefit of symbiosis for the alga, and while several have been located, while all others are digested (Kodama and Fujishima proposed, including photoprotection and escape from viral pre- 2011, 2014). Even once established, complete darkness or chem- dation (Reisser et al. 1991; Summerer et al. 2009; Esteban, Fenchel ical inhibitors, both of which prevent Chlorella photosynthesis and Finlay 2010), this interaction proves that a stable, even and therefore stop the carbon supply to the host, lead to the sometimes obligate, symbiosis can evolve from exploitation. eventual loss of Chlorella symbionts, through either digestion Other similar symbioses also appear to be founded upon or egestion (Karakashian 1963; Kodama and Fujishima 2008). In exploitation. For example, for scleractinian corals and the addition, cell division of symbiotic Chlorella is tightly regulated dinoflagellate algae Symbiodinium there is evidence of asymme- and has been linked to host cytoplasmic streaming (Takahashi try in the fitness effects of symbiosis upon the partners. The et al. 2007). Furthermore, metabolic processes are believed to algal growth rate is reduced from a free-living doubling time actively influence the exchange process, for instance host2 Ca + of 3 days to a symbiotic doubling time of between 70 and 100 inhibits serine uptake into Chlorella and glucose increases the days (Wilkerson, Kobayashi and Muscatine 1988). Whereas hosts experience increased growth rates in symbiosis. Further support Sørensen et al. 5

for the idea that this association is exploitative is provided by endosymbiotic algae excreted more glycerol and sucrose, and the asymmetry of the nutrient exchange: whilst the algal sym- contained more photopigments than the ancestral clone (Ger- biont provides ∼95% of its photosynthate to the host, in return mond et al. 2013). The evolved free-living algae adapted to the they are kept in a nitrogen-starved state by the host (Smith free-living environment and outcompeted any endosymbiotic and Muscatine 1999; Dubinsky and Berman-Frank 2001). Simi- algae that escaped symbiosis. This suggests that a trade-off larly, studies on lichen symbioses and the partnership between between adaptation to the free-living versus the symbiotic envi- chemosynthetic bacteria and their invertebrate hosts have also ronment may frequently enforce interspecific cooperation and reported reduced symbiont growth rates in symbiosis compared thus stabilise symbiosis, and is conceptually similar to the trade- Downloaded from https://academic.oup.com/femsle/article-abstract/366/12/fnz148/5528313 by University of Sheffield user on 09 August 2019 to free-living (Ahmadjian 1993; Combes 2005). Additionally, the off proposed by Law and Dieckmann (1998). association of Acantharia marine protists with haptophyte algae Although additional experimental evolution studies are is also believed to be a form of farming, whereby only the host clearly needed, it is intriguing that both studies to date support benefits (Decelle 2013). What these interactions have in com- the role for exploitation in the establishment of symbioses that mon is that they feature a producer living within a consumer. In evolve become mutualistic. Both experiments suggest a key role both the coral and P. bursaria symbioses, the algal symbionts are for trade-offs between symbiotic and free-living environments ‘engulfed’ during establishment and therefore do not actively in driving the emergence of mutualistic symbiosis, as predicted enter symbiosis. In symbiosis, the algae are contained within by Law and Dieckmann (1998). These experiments were essen- a host membrane, enabling the host to control provisioning of tially observational in design, lacking treatments to compare the resources. This inequality of control may be a defining feature of effects of environmental variables. Experiments manipulating apparently mutualistic symbioses founded upon exploitation. key environmental parameters likely to affect symbiosis, such as the potential for horizontal transmission or the free-living mor- Experimental evolution of microbial symbioses tality rate, will be an important next step towards understanding the environmental drivers of the establishment of symbiosis. Experimental evolution provides an unparalleled window into evolutionary processes by allowing their observation in real time CONCLUSION from defined genetic and phenotypic starting points under controlled conditions in the laboratory. While simplified lab envi- Both the theoretical and empirical evidence support the role for ronments preclude direct comparisons to nature, they allow key parasitism or exploitation in the establishment of symbioses, variables to be separated from the myriad of confounding vari- and the later evolution of mutual benefit. Establishment through ables in the field, providing a way to unambiguously separate exploitation provides a simple explanation for the establish- the proximate and ultimate causes of symbiosis (Mazancourt, ment of symbiosis because it does not require complex stabil- Loreau and Dieckmann 2005). ising mechanisms to repress conflict. Exploitation may be espe- To date there are only few examples of experimentally cially prevalent among associations where the smaller partner evolved establishments of novel symbiotic relationships. Jeon is engulfed by a larger host and enclosed in the host membrane. (1972) reported the first instance of an intracellular obligate par- In such associations, it is clear from the available experimental asite evolving to become a mutualistic symbiont. The exper- data that the core nutrient exchange between partners does not iment used Amoeba discoides that had become spontaneously in itself provide mutual benefits. It is likely that fitness trade- infected with rod-shaped bacteria and these were then cul- offs between the symbiotic and free-living environments play a tured together, without any selection for symbiosis, for five key role in enforcing exploitative symbioses, and may lead to the years. At first, the bacteria were harmful; the infected amoe- eventual emergence of dependence and mutual benefit through bae grew slower, were more sensitive to starvation, were smaller the loss of fitness in the free-living state. and some hosts cells were killed upon infection. However, after five years, the infected amoebae grew normally despite carrying the same number of bacteria cells. Crucially, this was not due ACKNOWLEDGEMENTS simply to the evolution of reduced virulence by the bacterium. AJW is grateful to Thorunn Helgason and Elva Robinson for stim- Nuclear transfer experiments swapped the evolved nucleus and ulating discussions. cytoplasm with that of the ancestor and demonstrated that the evolved nucleus could now not survive without the coevolved bacterial symbiont. Thus, a mutualistic and obligate symbiosis FUNDING had evolved from a parasitism. This work was funded by grant NE/K011774/2 from the Natu- More recently, Nakajima et al. (2009, 2015) established long- ral Environment Research Council, UK to MAB, CDL, DDC, and term microcosms containing a green alga (Micractinium sp., for- AJW, and a White Rose DTP studentship from the Biotech- mally Chlorella vulgaris), a bacterium (Escherichia coli), and a cil- nology and Biological Sciences Research Council, UK to MESS iate (Tetrahymena thermophila). The experiment was maintained (BB/M011151/1). The funders had no role in the design of the without external addition of resources and without transfer to study, the collection, analysis and interpretation of data, or the fresh medium for over five years and therefore formed a self- writing of the manuscript. sustaining ecosystem. Over the course of the experiment the free-living algae diversified into two distinct forms. One of these Conflicts of interest. None declared was a non-aggregating type that formed an endosymbiotic association with Tetrahymena as its host, whereas an aggregate form- REFERENCES ing type lived outside of Tetrahymena cells but formed a symbiotic association with the E. coli. The algal aggregation phenotype Ahmadjian V. The Lichen Symbiosis. John Wiley & Sons, 1993. was negatively correlated with Tetrahymena longevity in cocul- Albers D, Reisser W, Wiessner W. Studies on the nitrogen supply ture, suggesting that only non-aggregating algae improved host of endosymbiotic chlorellae in greem paramecium bursaria. fitness. Potentially underpinning this host benefit, the evolved Plant Sci Lett 1982;25:85–90. 6 FEMS Microbiology Letters, 2019, Vol. 366, No. 12

Archetti M, Scheuring I, Hoffman M et al. Economic game theory Jeon KW. Development of cellular dependence on infec- for mutualism and cooperation. Ecol Lett 2011;14:1300–12. tive organisms: micrurgical studies in amoebas. Science Axelrod R. The Evolution of Cooperation. Basic Books, 1984. 1972;176:1122–3. Bronstein JL ed. Mutualism. Oxford, New York: Oxford University Johnson MD. The acquisition of phototrophy: adaptive strate- Press, 2015 gies of hosting endosymbionts and organelles. Photosynth Res Bull JJ, Rice WR. Distinguishing mechanisms for the evolution of 2011;107:117–32. Co-Operation. J Theor Biol 1991;149:63–74. Kaltenpoth M, Roeser-Mueller K, Koehler S et al. Partner choice Combes C. The Art of Being a Parasite. University of Chicago Press, and fidelity stabilize coevolution in a cretaceous-age defen- Downloaded from https://academic.oup.com/femsle/article-abstract/366/12/fnz148/5528313 by University of Sheffield user on 09 August 2019 2005. sive symbiosis. Proc Natl Acad Sci 2014;111:6359–64. Currie CR, Scott JA, Summerbell RC et al. Fungus-growing ants Karakashian SJ. Growth of paramecium bursaria as influenced use antibiotic-producing bacteria to control garden para- by the presence of algal symbionts. Physiol Zool 1963;36:52– sites. Nature 1999;398:701–4. 68. Dean AD, Minter EJA, Sørensen MES et al. Host control and Kato Y, Imamura N. Effect of calcium ion on uptake of amino nutrient trading in a photosynthetic symbiosis. J Theor Biol acids by symbiotic chlorella F36-ZK isolated from Japanese 2016;405:82–93. paramecium bursaria. Plant Sci 2008a;174:88–96. De Bary A. The phenomenon of symbiosis. Strasbourg, Germany: Kato Y, Imamura N. Effect of sugars on amino acid transport by Karl J. Trubner, 1879. symbiotic chlorella. Plant Physiol Biochem 2008b;46:911–7. Decelle J. New perspectives on the functioning and evolu- Kato Y, Ueno S, Imamura N. Studies on the nitrogen utilization tion of photosymbiosis in plankton. Commun Integr Biol of endosymbiotic algae isolated from japanese paramecium 2013;6:e24560. bursaria. Plant Sci 2006;170:481–6. Dubinsky Z, Berman-Frank I. Uncoupling primary production Keeling PJ, McCutcheon JP. Endosymbiosis: The feeling is not from population growth in photosynthesizing organisms in mutual. J Theor Biol 2017;434:75–9. aquatic ecosystems. Aquat Sci 2001;63:4–17. Kiers ET, Rousseau RA, West SA et al. Host sanctions and the Esteban GF, Fenchel T, Finlay BJ. Mixotrophy in ciliates. Protist legume–rhizobium mutualism. Nature 2003;425:78–81. 2010;161: 621–41. Kiers ET, West SA. Evolving new organisms via symbiosis. Science Estrela S, Kerr B, Morris JJ. Transitions in individuality through 2015;348:392–4. symbiosis. Curr Opin Microbiol 2016;31:191–8. King KC, Brockhurst MA, Vasieva O et al. Rapid evolution of Fisher RM, Henry LM, Cornwallis CK et al. The evolu- microbe-mediated protection against pathogens in a worm tion of host-symbiont dependence. Nature Commun host. ISME J 2016;10:1915–24. 2017;8:ncomms15973. Kodama Y, Fujishima M. Cell division and density of symbiotic Frank SA. Models of symbiosis. Am Nat 1997;150: S80–99. chlorella variabilis of the ciliate paramecium bursaria is con- Frank SA. Policing and group cohesion when resources vary. trolled by the host’s nutritional conditions during early infec- Anim Behav 1996;52:1163–9. tion process. Environ Microbiol 2012;14:2800–11. Frank SA. The origin of synergistic symbiosis. J Theor Biol Kodama Y, Fujishima M. Cycloheximide induces synchronous 1995;176:403–10. swelling of perialgal vacuoles enclosing symbiotic chlorella Frederickson ME. Rethinking mutualism stability: cheaters and vulgaris and digestion of the algae in the ciliate paramecium the evolution of sanctions. Q Rev Biol 2013;88:269–95. bursaria. Protist 2008;159:483–94. Garcia JR, Gerardo NM. The symbiont side of symbiosis: do Kodama Y, Fujishima M. Four important cytological events microbes really benefit? Fronti Microbiol 2014;5:510. needed to establish endosymbiosis of symbiotic Chlorella Genkai-Kato M, Yamamura N. Evolution of mutualistic Sp. to the alga-free paramecium bursaria. Japan. J Protozool symbiosis without vertical transmission. Theor Popul Biol 2011;44:1–20. 1999;55:309–23. Kodama Y, Fujishima M. Symbiotic chlorella variabilis incubated Germond A, Kunihiro T, Inouhe M et al. Physiological changes under constant dark conditions for 24 hours loses the ability of a green alga (Micractinium Sp.) involved in an early-stage to avoid digestion by host lysosomal enzymes in digestive of association with tetrahymena thermophila during 5-Year vacuoles of host ciliate paramecium bursaria. FEMS Microbiol microcosm culture. Biosystems 2013;114:164–71. Ecol 2014;90:946–55. Hamilton WD. The genetical evolution of social behaviour. II. J Law R, Dieckmann U. Symbiosis through exploitation and the Theor Biol 1964;7:17–52. merger of lineages in evolution. Proc Roy Soc Lon B: Biol Sci Hardin G. The tragedy of the commons. Science 1968;162:1243–8. 1998;265:1245–53. Heath KD, Tiffin P. Context dependence in the coevolution of Leigh EG, Jr. The evolution of mutualism. J Evol Biol 2010;23:2507– plant and rhizobial mutualists. Proce Royal Soc Lond B: Biol Sci 28. 2007;274:1905–12. Lowe CD, Minter EJ, Cameron DD et al. Shining a light on exploita- Herre EA, Knowlton N, Mueller UG et al. The Evolution of Mutu- tive host control in a photosynthetic endosymbiosis. Curr Biol alisms: Exploring the Paths between Conflict and Coopera- 2016;26:207–11. tion.” Trends Ecol Evol 1999;14:49–53. Mazancourt CD, Loreau M, Dieckmann U. Understanding mutu- Hoang KL, Morran LT, Gerardo NM. Experimental evolution as alism when there is adaptation to the partner. J Ecol an underutilized tool for studying beneficial animal–microbe 2005;93:305–14. interactions. Front Microbiol 2016;7.1444. McFall-Ngai MJ, Ruby EG. Symbiont recognition and subsequent Hoshina R, Imamura N. Multiple origins of the symbioses in morphogenesis as early events in an animal-bacterial mutu- paramecium bursaria. Protist 2008;159:53–63. alism. Science 1991;254:1491–4. Hortnagl¨ PH, Sommaruga R. Photo-oxidative stress in symbiotic Minter EJA, Lowe CD, Sørensen MES et al. Variation and asymme- and aposymbiotic strains of the ciliate paramecium bursaria. try in host-symbiont dependence in a microbial symbiosis. Photochem Photobiol Sci 2007;6:842. BMC Evol Biol 2018;18:108. Sørensen et al. 7

Moran NA, Plague GR, Sandstrom¨ JP et al. A genomic perspective Soldo AT, Godoy GA, Larin F. Purine-Excretory nature of refractile on nutrient provisioning by bacterial symbionts of insects. bodies in the marine ciliate parauronema acutum∗. J Protozool Proc Natl Acad Sci 2003;100:14543–8. 1978;25:416–8. Muscatine L, Porter JW. Reef corals: Mutualistic sym- Sprent JI, Sutherland JM, De Faria SM. Some aspects of the bioses adapted to nutrient-poor environments. Bioscience biology of nitrogen-fixing organisms. Phil Trans R Soc Lond B 1977;27:454–60. 1987;317:111–29. Nakajima T, Fujikawa Y, Matsubara T et al. Differentiation of Summerer M, Sonntag B, Hortnagl¨ P et al. Symbiotic ciliates a free-living alga into forms with ecto- and endosymbiotic receive protection against UV damage from their algae: Downloaded from https://academic.oup.com/femsle/article-abstract/366/12/fnz148/5528313 by University of Sheffield user on 09 August 2019 associations with heterotrophic organisms in a 5-year micro- A test with paramecium bursaria and chlorella. Protist cosm culture. Biosystems 2015;131:9–21. 2009;160:233–43. Nakajima T, Sano A, Matsuoka H. Auto-/Heterotrophic Summerer M, Sonntag B, Sommaruga R. Ciliate-symbiont speci- endosymbiosis evolves in a mature stage of ecosystem ficity of freshwater endosymbiotic chlorella (Trebouxio- development in a microcosm composed of an alga, a phyceae, Chlorophyta)1. Journal of Phycology 2008;44:77–84. bacterium and a ciliate. Biosystems 2009;96:127–35. Szathmary´ E, Smith JM. The major evolutionary transitions. Nussbaumer AD, Fisher CR, Bright M. Horizontal endosym- Nature 1995;374:227–32. biont transmission in hydrothermal vent tubeworms. Nature Takahashi T, Shirai Y, Kosaka T et al. Arrest of cytoplasmic 2006;441:345–8. streaming induces algal proliferation in green paramecia. Perez S, Weis V. Nitric oxide and cnidarian bleaching: An evic- PLoS One 2007;2:e1352. tion notice mediates breakdown of a symbiosis. J Exp Biol Tebo BM, Linthicum DS, Nealson KH. Luminous bacteria and 2006;209:2804–10. light emitting fish: ultrastructure of the symbiosis. Biosystems Pfeffer PE, Douds DD, Becard´ G et al. Carbon uptake and the 1979;11:269–80. metabolism and transport of lipids in an arbuscular mycor- Tso GHW, Reales-Calderon JA, Tan ASM et al. Experimental evo- rhiza. Plant Physiol 1999;120:587–98. lution of a fungal pathogen into a gut symbiont. Science Powell JR, Rillig MC. Biodiversity of arbuscular mycorrhizal fungi 2018;362:589–95. and ecosystem function. New Phytol 2018;220:1059–75. Tsuchida T, Koga R, Horikawa M et al. Symbiotic bacterium mod- Rankin DJ, Bargum K, Kokko H. The tragedy of the commons in ifies aphid body color. Science 2010;330:1102–4. evolutionary biology. Trends Ecol Evol 2007;22:643–51. Visick KL, Ruby EG. Vibrio fischeri and its host: it takes twoto Reisser W, Burbank D, Meints R et al. Viruses distinguish symbi- tango. Curr Opin Microbiol 2006;9:632–8. otic chlorella Spp of paramecium-bursaria. Endocytobiosis Cell Wendling CC, Fabritzek AG, Wegner KM. Population-specific Res 1991;7:245–51. genotype x genotype x environment interactions in bacterial Sachs JL, Mueller UG, Wilcox TP et al. The evolution of coopera- disease of early life stages of pacific oyster larvae. Evol Appl tion. Q Rev Biol 2004;79:135–60. 2017;10:338–47. Sachs JL, Simms EL. Pathways to mutualism breakdown. Trends Wernegreen JJ. Endosymbiosis: Lessons in conflict resolution. Ecol Evol 2006;21:585–92. PLOS Biol 2004;2:e68. Sachs JL, Skophammer RG, Regus JU. Evolutionary transitions in Wilkerson FP, Kobayashi D, Muscatine L. Mitotic index and bacterial symbiosis. Proc Natl Acad Sci 2011;108:10800–807. size of symbiotic algae in caribbean reef corals. Coral Reefs Shah N, Syrett PJ. The uptake of guanine and hypoxanthine by 1988;7:29–36. marine microalgae. J Mar Biol Assoc UK 1984;64:545–56. Yamamura N. Vertical transmission and evolution of mutualism Shapiro JW, Turner PE. Evolution of mutualism from parasitism from parasitism. Theor Popul Biol 1993;44:95–109. in experimental virus populations. Evolution 2018;72:707–12. Ziesenisz E, Reisser W, Wiessner W. Evidence of de novo synthe- Smith GJ, Muscatine L. Cell cycle of symbiotic dinoflagellates: sis of maltose excreted by the endosymbiotic chlorella from Variation in G1 phase-duration with anemone nutritional paramecium bursaria. Planta 1981;153:481–5. status and macronutrient supply in the aiptasia pulchella– symbiodinium pulchrorum symbiosis. Mar Biol 1999;134:405– 18. Appendix B Statistical outputs for Chapter 2. The analyses associated with Figures 2.2 and 2.5. In the majority of cases, responses were analysed as ANOVA models. For ΦPSII responses reported in Figure 2.5b, a non-linear mixed effects model was used.

Relating to Figure 2.2 ANOVA model for selection rate in response to growth irradiance analysed by host genotype (following model reduction)

Factor DF SS MSS F value p value Host 1 0.516 0.516 19.387 <0.001 Growth irradiance 1 1.6963 1.6963 63.731 <0.001 Host:Growth irradiance 1 0.1308 0.1308 4.915 0.034088 Residuals 31 0.8251 0.0266 F-statistic: 29.34 on 3 and 31 DF, p-value: 3.469e-09, Adjusted R2: 0.7144

Relating to Figure 2.5a ANOVA model for FvFm estimates in response to light analysed by host and symbiont genotype

Factor DF SS MSS F value p value Host 1 0.07719 0.07719 121.099 <0.001 Symbiont 1 0.00006 0.00006 0.088 0.767 Light 1 0.00418 0.00418 6.563 0.011 Host:Symbiont 1 0.03891 0.03891 61.052 <0.001 Host:Light 1 0.01305 0.01305 20.481 <0.001 Symbiont:Light 1 0.12213 0.12213 191.616 <0.001 Host:Symbiont:Light 1 0.13001 0.13001 203.984 <0.001 Residuals 232 0.14787 0.00064 F-statistic: 86.41 on 7 and 232 DF, p-value: < 2.2e-16, Adjusted R2: 0.7144

Tukey HSD posthoc test, showing the result for symbiont comparison (18-HA): Host Light Difference P adj 186 12 0.067 0.000 50 -0.116 0.000 HA1 12 0.025 0.004 50 0.028 0.001

Relating to Figure 2.5b Non-linear mixed effects model (assuming exponential decay of the form y = a × e(light×b)) for the response of steady-state quantum yield (ΦPSII) to actinic light analysed by growth irradiance, host genotype and symbiont genotype. Replicates within treatments were treated as random effects; growth irradiance, host identity and symbiont identity were treated as fixed effects.

Model DF AIC BIC logLik Test L.Ratio p-value 1)Host*Symbiont*Light 17 -11478.35 -11382.7 5756.176 2)Host*Symbiont 9 -11446.53 -11395.9 5732.268 1vs2 47.81662 <.0001 3)Host*Light 11 -11449.02 -11387.1 5735.508 1vs3 41.33577 <.0001 4)Symbiont*Light 11 -11452.35 -11390.5 5737.174 1v4 38.00399 <.0001

131

Estimates of coefficients Symbiont Light Intercept SE exponent SE Host 186 18 12 0.342515 0.01121 -0.001247 0.0000229 24 0.419108 0.011165 -0.001247 0.0000229 50 0.409108 0.011069 -0.001247 0.0000229 HA 12 0.407578 0.011079 -0.001247 0.0000229 24 0.40278 0.011093 -0.001247 0.0000229 50 0.304538 0.011219 -0.001247 0.0000229 HA1 18 12 0.368342 0.011084 -0.001247 0.0000229 24 0.427061 0.011054 -0.001247 0.0000229 50 0.39956 0.011122 -0.001247 0.0000229 HA 12 0.416999 0.011122 -0.001247 0.0000229 24 0.421092 0.0111 -0.001247 0.0000229 50 0.426505 0.011084 -0.001247 0.0000229

ANOVA on the summary statistics - the response of the nlme predicted intercept to experimental group (a single factor that combines host genotype, symbiont genotype and growth irradiance). Computed by aovSufficient (HH package)

Factor DF SS MSS F value p value Group 11 0.04755 0.004323 11.66 <0.001 Residuals 24 0.0089 0.000371 F-statistic: 11.66 on 11 and 24 DF, p-value: 2.28e-07

Tukey HSD posthoc test, showing the results for the symbiont comparison (18-HA): Group Host Light Difference P adj 186 12 0.065 0.015 24 -0.016 0.995 50 -0.105 0.000 HA1 12 0.049 0.142 24 -0.006 1.000 50 0.027 0.845

Relating to Figure 2.5c NSV values are modelled by polynomial models in the form Y = ax2 +bx + c. Each coefficient was then evaluated by ANOVA models to test which factors significantly affect them.

Coefficient c (the intercept) - linear model for its response to host genotype (following model reduction) Factor DF SS MSS F value p value Host 1 1.15 1.15 4.739 0.0365 Residuals 34 8.251 0.2427 F-statistic: 4.739 on 1 and 34 DF, p-value: 0.03653, Adjusted R2: 0.09651

132

Coefficient b - linear model for its response to symbiont genotype and growth irradiance (following model reduction) Factor DF SS MSS F value p value Symbiont 1 1.52E-06 1.52E-06 9.485 <0.01 Growth Irradiance 2 1.16E-06 5.78E-07 3.6 0.03888 Residuals 32 5.14E-06 1.61E-07 F-statistic: 5.562 on 3 and 32 DF, p-value: 0.003456, Adjusted R2: 0.2811

Coefficient a - linear model for its response to symbiont genotype (following model reduction) Factor DF SS MSS F value p value Symbiont 1 1.07E-13 1.07E-13 8.932 <0.01 Residuals 34 4.08E-13 1.20E-14 F-statistic: 8.932 on 1 and 34 DF, p-value: 0.005176, Adjusted R2: 0.1847

133

Appendix C Statistical outputs for Chapter 3. The analyses associated with Figures 3.2,3.3,3.8. In most cases, the responses were analysed with ANOVA models.

Relating to Figure 3.2 ANOVA on growth rates in response to growth iradiance analysed by host and symbiont genotype Factor DF SS MSS F value p value Host 2 2.79E-02 1.39E-02 2.262 0.10738 Symbiont 2 7.76E-02 3.88E-02 6.297 <0.01 Light 1 1.49E+00 1.49E+00 242.246 <0.001 Host:Symbiont 4 8.24E-02 2.06E-02 3.345 0.0116 Host:Light 2 6.26E-02 3.13E-02 5.084 <0.01 Symbiont:Light 2 1.38E-01 6.91E-02 11.215 <0.001 Host:Symbionts:Light 4 8.87E-02 2.22E-02 3.6 <0.01 Residuals 162 9.98E-01 6.20E-03 2 F-statistic: 18.81 on 17 and 162 DF, p-value: <0.001, Adjusted R : 0.63

Relating to Figure 3.3 ANOVA on symbiont load in response to growth irradiance analysed by host and symbiont genotype Factor DF SS MSS F value p value Host 2 5.75E+12 2.87E+12 10.869 <0.001 Symbiont 2 2.00E+12 1.00E+12 3.787 0.0247 Light 1 1.56E+12 1.56E+12 5.911 0.0161 Host:Symbiont 4 1.47E+12 3.67E+11 1.388 0.2404 Host:Light 2 1.64E+12 8.20E+11 3.101 0.0477 Symbiont:Light 2 1.65E+12 8.25E+11 3.12 0.0468 Host:Symbiont:Light 4 2.94E+12 7.35E+11 2.78 0.0287 Residuals 162 4.28E+13 2.64E+11 F-statistic: 3.784 on 17 and 162 DF, p-value: <0.001, Adjusted R2: 0.2091

Symbiont load values were modelled by polynomial models in the form Y = ax2 +bx + c. Each coefficient was then evaluated by linear models to test which factors (host and symbiont genotype) significantly affect them.

Coefficient c (the intercept) - linear model for its response to host and symbiont genotype Factor DF SS MSS F value p value host 2 6.18E+12 3.09E+12 58.47 <0.001 symbiont 2 2.20E+12 1.10E+12 20.82 <0.001 host:symbiont 4 3.13E+12 7.81E+11 14.79 <0.001 Residuals 36 1.90E+12 5.28E+10 F-statistic: 27.22 on 8 and 36 DF, p-value: <0.001 , Adjusted R2: 0.83

134

Coefficient b (first coefficient) - linear model for its response to host and symbiont genotype Factor DF SS MSS F value p value host 2 2.26E+10 1.13E+10 17.987 <0.001 symbiont 2 6.92E+09 3.46E+09 5.503 <0.01 host:symbiont 4 1.36E+10 3.41E+09 5.413 <0.01 Residuals 36 2.26E+10 6.29E+08 F-statistic: 8.58 on 8 and 36 DF, p-value: <0.001 , Adjusted R2: 0.58

Coefficient a (second coefficient) - linear model for its response to host and symbiont genotype Factor DF SS MSS F value p value host 2 1.56E+12 7.78E+11 13.836 <0.001 symbiont 2 4.57E+11 2.28E+11 4.063 0.0257 host:symbiont 4 7.28E+11 1.82E+11 3.236 0.0229 Residuals 36 2.02E+12 5.62E+10 F-statistic: 6.09 on 8 and 36 DF, p-value: <0.001 , Adjusted R2: 0.48

The polynomial models were used to calculate predictive values for the coordinates at the peak maximum.

X max - ANOVA on the predicted X values at the peak maximum in response to host and symbiont genotype Factor DF SS MSS F value p value host 2 165.9 82.96 9.634 <0.001 symbiont 2 192.8 96.42 11.197 <0.001 host:symbiont 4 474 118.49 13.759 <0.001 Residuals 36 310 8.61 F-statistic: 12.09 on 8 and 36 DF, p-value: <0.001 , Adjusted R2: 0.67

Y max - ANOVA on the predicted Y values at the peak maximum in response to host and symbiont genotype Factor DF SS MSS F value p value host 2 4.42E+11 2.21E+11 5.596 <0.01 symbiont 2 7.88E+11 3.94E+11 9.97 <0.001 host:symbiont 4 6.03E+11 1.51E+11 3.812 0.011 Residuals 36 1.42E+12 3.95E+10 F-statistic: 5.80 on 8 and 36 DF, p-value: <0.001 , Adjusted R2: 0.47

135

Tukey HSD posthoc tests summary table for the intercept, X maximum and Y maximum values. Showing the result for symbiont comparisons: Intercept X-max Y-max Host Pairwise Tests Difference P adj Difference P adj Difference P adj HA1 h:HA - h:18 -98859 0.999 4.429 0.322 -197462 0.814 h:HK - h:18 -192634 0.917 3.529 0.617 -13529 1.000 h:HK - h:HA -93775 0.999 -0.900 1.000 183933 0.865 HK1 k:HA - k:18 690394 0.001 -3.335 0.684 -52652 1.000 k:HK - k:18 512444 0.028 -3.379 0.669 534623 0.004 k:HK - k:HA -177949 0.946 -0.044 1.000 587275 0.001 186 s:HA - s:18 546447 0.016 -5.764 0.078 -284314 0.391 s:HK - s:18 1253004 0.000 -15.023 0.000 -84741 0.999 s:HK - s:HA 706557 0.001 -9.259 0.000 199573 0.805

Relating to Figure 3.8 ANOVA on the mz 686.4 relative abundance in response to host and symbiont genotype Factor DF SS MSS F value p value host 2 9.24E-07 4.62E-07 0.396 0.67852 symbiont 2 1.29E-05 6.47E-06 5.55 0.01326 host:symbiont 4 2.16E-05 5.40E-06 4.632 <0.01 Residuals 18 2.10E-05 1.17E-06 F-statistic: 13.802 on 8 and 18 DF, p-value: 0.008859 , Adjusted R2: 0.463

ANOVA on the mz 271.2 relative abundance in response to host and symbiont genotype Factor DF SS MSS F value p value host 2 3.81E-06 1.91E-06 0.901 0.4237 symbiont 2 3.08E-06 1.54E-06 0.728 0.4966 host:symbiont 4 2.94E-05 7.34E-06 3.468 0.0287 Residuals 18 3.81E-05 2.12E-06 F-statistic: 2.141 on 8 and 18 DF, p-value: 0.08574 , Adjusted R2: 0.2599

ANOVA on the mz 247.2 relative abundance in response to host and symbiont genotype Factor DF SS MSS F value p value host 2 1.13E-06 5.64E-07 0.852 0.443 symbiont 2 1.85E-06 9.24E-07 1.394 0.274 host:symbiont 4 6.00E-06 1.50E-06 2.264 0.102 Residuals 18 1.19E-05 6.63E-07 F-statistic: 1.693 on 8 and 18 DF, p-value: 0.1681 , Adjusted R2: 0.1758

136

Appendix D Statistical outputs for Chapter 4. Analyses associated with Figures 4.1 - 4.4. In most cases responses were analysed as ANOVA models. For the growth rate per transfer (Fig 4.1), a linear mixed effects model was used.

Relating to Figure 4.1 Linear mixed effect model for the response of growth rate to transfer week analysed by symbiont genotype. Transfer week within treatment ID were treated as random effects; symbiont genotype and transfer week were treated as fixed effects.

Model DF AIC BIC logLik Test L.Ratio p-value 1) Symbiont + Transfer 7 -1123.768 -1098.13 568.8839 2) Symbiont 6 -1118.694 -1096.72 565.3468 1v2 7.074102 0.0078 3) Transfer 6 -1094.306 -1072.33 553.1531 1v3 31.46149 <.0001

Fixed effects Estimate SE DF T-value Intercept 0.281 0.007 275 42.266 SymbiontSHK -0.080 0.006 10 -14.126 transfer 0.001 0.000 275 3.088

Random effects SD Correlation Intercept 0.015 Transfer 0.001 -0.878 Residual 0.033

Relating to Figure 4.2 ANOVA for growth assay in response to light analysed by symbiont genotype and transfer number Factor DF SS MSS F value p value Light 3 2.133 0.711 216.332 <0.001 Symbiont 1 0.0424 0.0424 12.911 <0.001 Transfer 3 0.0568 0.0189 5.766 <0.001 Light:Symbiont 3 0.0758 0.0253 7.688 <0.001 Symbiont:Transfer 3 0.0908 0.0303 9.204 <0.001 Residuals 178 0.585 0.0033 F-statistic: 56.14 on 13 and 178 DF, p-value: < 2.2e-16, Adjusted R2: 0.7896

Relating to Figure 4.3 ANOVA for symbiont-load in response to light analysed by symbiont genotype and transfer n umber Factor DF SS MSS F value p value Symbiont 1 1.85E+12 1.85E+12 18.167 <0.001 Light 4 4.47E+13 1.12E+13 109.564 <0.001 Transfer 1 5.37E+12 5.37E+12 52.578 <0.001 Symbiont:Light 4 4.55E+12 1.14E+12 11.158 <0.001 Symbiont:Transfer 1 2.79E+12 2.79E+12 27.366 <0.001 Light:Transfer 4 4.60E+12 1.15E+12 11.257 <0.001 Symbiont:Light:Transfer 4 2.32E+12 5.81E+11 5.693 <0.001 Residuals 76 7.76E+12 1.02E+11 F-statistic: 34.15 on 19 and 76 DF, p-value: < 2.2e-16, Adjusted R2: 0.8689 137

Relating to Figure 4.4 ANOVA for selection rate in response to light analysed by symbiont genotype and transfer number Factor DF SS MSS F value p value Symbiont 1 0.1293 0.1293 7.484 <0.01 Light 2 0.6747 0.3373 19.527 <0.001 Transfer 1 0.511 0.511 29.58 <0.001 Symbiont:Light 2 0.0484 0.0242 1.401 0.258 Symbiont:Transfer 1 0.0117 0.0117 0.679 0.415 Light:Transfer 2 0.1075 0.0537 3.111 0.055 Symbiont:Light:Transfer 2 0.2032 0.1016 5.882 <0.01 Residuals 41 F-statistic: 8.871 on 11 and 41 DF, p-value: 8.663e-08, Adjusted R2: 0.6248

Tukey HSD posthoc test, showing the result for symbiont comparison (18-HK): Transfer Light Difference P adj T0 0 0.134 0.992 12 -0.092 0.999 50 -0.395 0.029 T25 0 -0.176 0.483 12 0.022 1.000 50 -0.046 1.000

138

Bibliography Abdel-Rahman, M.H.M., Ali, R.M., and Said, H.A. (2005). Alleviation of NaCl-induced effects on Chlorella vulgaris andChlorococcum humicola by riboflavin application. International Journal of Agriculture and Biology (Pakistan).

Abrego, D., Ulstrup, K.E., Willis, B.L., and van Oppen, M.J.H. (2008). Species–specific interactions between algal endosymbionts and coral hosts define their bleaching response to heat and light stress. Proceedings of the Royal Society B: Biological Sciences 275, 2273–2282.

Achlatis, M., Pernice, M., Green, K., Guagliardo, P., Kilburn, M.R., Hoegh-Guldberg, O., and Dove, S. (2018). Single-cell measurement of ammonium and bicarbonate uptake within a photosymbiotic bioeroding sponge. ISME J 12, 1308–1318.

Ahmadjian, V. (1993). The Lichen Symbiosis (John Wiley & Sons).

Albers, D., Reisser, W., and Wiessner, W. (1982). Studies on the nitrogen supply of endosymbiotic chlorellae in greem paramecium bursaria. Plant Science Letters 25, 85–90.

Allen, J.F., Raven, J.A., Embley, T.M., der Giezen Mark van, Horner David S., Dyal Patricia L., and Foster Peter (2003). Mitochondria and hydrogenosomes are two forms of the same fundamental organelle. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences 358, 191–203.

Anbutsu, H., Moriyama, M., Nikoh, N., Hosokawa, T., Futahashi, R., Tanahashi, M., Meng, X.-Y., Kuriwada, T., Mori, N., Oshima, K., et al. (2017). Small genome symbiont underlies cuticle hardness in beetles. PNAS 114, E8382–E8391.

Ankrah, N.Y.D., and Douglas, A.E. (2018). Nutrient factories: metabolic function of beneficial microorganisms associated with insects. Environmental Microbiology 20, 2002–2011.

Anthony, K., Bay, L.K., Costanza, R., Firn, J., Gunn, J., Harrison, P., Heyward, A., Lundgren, P., Mead, D., Moore, T., et al. (2017). New interventions are needed to save coral reefs. Nat Ecol Evol 1, 1420–1422.

Archetti, M., Scheuring, I., Hoffman, M., Frederickson, M.E., Pierce, N.E., and Yu, D.W. (2011). Economic game theory for mutualism and cooperation. Ecology Letters 14, 1300– 1312.

Archibald, J.M. (2009). The Puzzle of Plastid Evolution. Current Biology 19, R81–R88.

Archibald, J.M., and Keeling, P.J. (2002). Recycled plastids: a ‘green movement’ in eukaryotic evolution. Trends in Genetics 18, 577–584.

Arnow, P., Oleson, J.J., and Williams, J.H. (1953). The Effect of Arginine on the Nutrition of Chlorella vulgaris. American Journal of Botany 40, 100–104.

Asensi-Fabado, M.A., and Munné-Bosch, S. (2010). Vitamins in plants: occurrence, biosynthesis and antioxidant function. Trends in Plant Science 15, 582–592.

139

Baird, A.H., Bhagooli, R., Ralph, P.J., and Takahashi, S. (2009). Coral bleaching: the role of the host. Trends in Ecology & Evolution 24, 16–20.

Baker, A.C. (2003). Flexibility and Specificity in Coral-Algal Symbiosis: Diversity, Ecology, and Biogeography of Symbiodinium. Annual Review of Ecology, Evolution, and Systematics 34, 661–689.

Barbrook, A.C., Howe, C.J., and Purton, S. (2006). Why are plastid genomes retained in non-photosynthetic organisms? 11, 101–108.

Bartosz, G. (1997). Oxidative stress in plants. Acta Physiol Plant 19, 47–64.

Bell-Roberts, L., Douglas, A.E., and Werner, G.D.A. (2019). Match and mismatch between dietary switches and microbial partners in plant sap-feeding insects. Proceedings of the Royal Society B: Biological Sciences 286, 20190065.

Bennett, G.M., and Moran, N.A. (2015). Heritable symbiosis: The advantages and perils of an evolutionary rabbit hole. PNAS 112, 10169–10176.

Benton, H.P., Want, E.J., and Ebbels, T.M.D. (2010). Correction of mass calibration gaps in liquid chromatography-mass spectrometry metabolomics data. Bioinformatics 26, 2488–2489.

Berkelmans, R., and van Oppen, M.J.H. (2006). The role of zooxanthellae in the thermal tolerance of corals: a ‘nugget of hope’ for coral reefs in an era of climate change. Proceedings of the Royal Society B: Biological Sciences 273, 2305–2312.

Berney, C., and Pawlowski, J. (2006). A molecular time-scale for eukaryote evolution recalibrated with the continuous microfossil record. Proceedings of the Royal Society B: Biological Sciences 273, 1867–1872.

Bhattacharya, D., and Archibald, J.M. (2006). Response to theissen and martin. Current Biology 16, R1017–R1018.

Blanc, G., Duncan, G., Agarkova, I., Borodovsky, M., Gurnon, J., Kuo, A., Lindquist, E., Lucas, S., Pangilinan, J., Polle, J., et al. (2010). The Chlorella variabilis NC64A Genome Reveals Adaptation to Photosymbiosis, Coevolution with Viruses, and Cryptic Sex. Plant Cell 22, 2943–2955.

Bodył, A., Mackiewicz, P., and Stiller, J.W. (2007). The intracellular cyanobacteria of Paulinella chromatophora: endosymbionts or organelles? Trends in Microbiology 15, 295–296.

Bonen, L., and Doolittle, W.F. (1975). On the prokaryotic nature of red algal chloroplasts. PNAS 72, 2310–2314.

Bonen, L., Cunningham, R.S., Gray, M.W., and Doolittle, W.F. (1977). Wheat embryo mitochondrial 18S ribosomal RNA: evidence for its prokaryotic nature. Nucleic Acids Res 4, 663–671.

Boulotte, N.M., Dalton, S.J., Carroll, A.G., Harrison, P.L., Putnam, H.M., Peplow, L.M., and van Oppen, M.J. (2016). Exploring the Symbiodinium rare biosphere provides

140 evidence for symbiont switching in reef-building corals. The ISME Journal 10, 2693– 2701.

Bourtzis, K., Dobson, S.L., Xi, Z., Rasgon, J.L., Calvitti, M., Moreira, L.A., Bossin, H.C., Moretti, R., Baton, L.A., Hughes, G.L., et al. (2014). Harnessing mosquito–Wolbachia symbiosis for vector and disease control. Acta Tropica 132, S150–S163.

Bronstein, J.L. (2015). Mutualism (Oxford University Press).

Buddemeier, R.W., and Fautin, D.G. (1993). Coral Bleaching as an Adaptive Mechanism. BioScience 43, 320–326.

Bull, J.J., and Rice, W.R. (1991). Distinguishing mechanisms for the evolution of cooperation. Journal of Theoretical Biology 149, 63–74.

Cameron, D.D., Leake, J.R., and Read, D.J. (2006). Mutualistic mycorrhiza in orchids: evidence from plant–fungus carbon and nitrogen transfers in the green-leaved terrestrial orchid Goodyera repens. New Phytologist 171, 405–416.

Cameron, D.D., Geniez, J.-M., Seel, W.E., and Irving, L.J. (2008). Suppression of Host Photosynthesis by the Parasitic Plant Rhinanthus minor. Ann Bot 101, 573–578.

Cantin, N.E., van Oppen, M.J.H., Willis, B.L., Mieog, J.C., and Negri, A.P. (2009). Juvenile corals can acquire more carbon from high-performance algal symbionts. Coral Reefs 28, 405.

Caragata, E.P., Dutra, H.L.C., and Moreira, L.A. (2016). Exploiting Intimate Relationships: Controlling Mosquito-Transmitted Disease with Wolbachia. Trends in Parasitology 32, 207–218.

Caspi, R., Billington, R., Fulcher, C.A., Keseler, I.M., Kothari, A., Krummenacker, M., Latendresse, M., Midford, P.E., Ong, Q., Ong, W.K., et al. (2018). The MetaCyc database of metabolic pathways and enzymes. Nucleic Acids Res 46, D633–D639.

Cavalier-Smith, T. (2013). Symbiogenesis: Mechanisms, Evolutionary Consequences, and Systematic Implications. Annual Review of Ecology, Evolution, and Systematics 44, 145–172.

Cavanaugh, C.M., Gardiner, S.L., Jones, M.L., Jannasch, H.W., and Waterbury, J.B. (1981). Prokaryotic Cells in the Hydrothermal Vent Tube Worm Riftia pachyptila Jones: Possible Chemoautotrophic Symbionts. Science 213, 340–342.

Chavez-Dozal, A., and Nishiguchi, M.K. (2016). Impact of Metabolomics in Symbiosis Research. Metabolomics: Fundamentals and Applications 139.

Chong, R.A., and Moran, N.A. (2016). Intraspecific genetic variation in hosts affects regulation of obligate heritable symbionts. PNAS 113, 13114–13119.

Coles, S.L., Bahr, K.D., Rodgers, K.S., May, S.L., McGowan, A.E., Tsang, A., Bumgarner, J., and Han, J.H. (2018). Evidence of acclimatization or adaptation in Hawaiian corals to higher ocean temperatures. PeerJ 6, e5347.

141

Combes, C. (2005). The Art of Being a Parasite (University of Chicago Press).

Converti, A., Casazza, A.A., Ortiz, E.Y., Perego, P., and Del Borghi, M. (2009). Effect of temperature and nitrogen concentration on the growth and lipid content of Nannochloropsis oculata and Chlorella vulgaris for biodiesel production. Chemical Engineering and Processing: Process Intensification 48, 1146–1151.

Corliss, J.O. (1961). The Ciliated Protozoa: Characterization, Classification and Guide to the Literature (Elsevier).

Corsaro, D., Venditti, D., Padula, M., and Valassina, M. (1999). Intracellular Life. Critical Reviews in Microbiology 25, 39–79.

Cunning, R., Vaughan, N., Gillette, P., Capo, T.R., Maté, J.L., and Baker, A.C. (2015). Dynamic regulation of partner abundance mediates response of reef coral symbioses to environmental change. Ecology 96, 1411–1420.

Currie, C.R., Scott, J.A., Summerbell, R.C., and Malloch, D. (1999). Fungus-growing ants use antibiotic-producing bacteria to control garden parasites. Nature 398, 701–704.

Dahan, R.A., Duncan, R.P., Wilson, A.C., and Dávalos, L.M. (2015). Amino acid transporter expansions associated with the evolution of obligate endosymbiosis in sap- feeding insects (Hemiptera: sternorrhyncha). BMC Evolutionary Biology 15, 52.

De Bary, A. (1879). The phenomenon of symbiosis. Karl J. Trubner, Strasbourg, Germany.

Dean, A.D., Minter, E.J.A., Sørensen, M.E.S., Lowe, C.D., Cameron, D.D., Brockhurst, M.A., and Jamie Wood, A. (2016). Host control and nutrient trading in a photosynthetic symbiosis. Journal of Theoretical Biology 405, 82–93.

Decelle, J. (2013). New perspectives on the functioning and evolution of photosymbiosis in plankton. Communicative & Integrative Biology 6, e24560.

Decelle, J., Colin, S., and Foster, R.A. (2015). Photosymbiosis in Marine Planktonic Protists. In Marine Protists, (Springer, Tokyo), pp. 465–500.

Delwiche, C.F. (1999). Tracing the Thread of Plastid Diversity through the Tapestry of Life. The American Naturalist 154, S164–S177.

Demirbas, M.F. (2011). Biofuels from algae for sustainable development. Applied Energy 88, 3473–3480.

DeSalvo, M.K., Estrada, A., Sunagawa, S., and Medina, M. (2012). Transcriptomic responses to darkness stress point to common coral bleaching mechanisms. Coral Reefs 31, 215–228.

Dodson, E.O. (1979). Crossing the prokaryote–eucaryote border: endosymbiosis or continuous development? Canadian Journal of Microbiology 25, 651–674.

Dohlen, C.D. von, Kohler, S., Alsop, S.T., and McManus, W.R. (2001). Mealybug β- proteobacterial endosymbionts contain γ-proteobacterial symbionts. Nature 412, 433.

142

Dorrell, R.G., and Smith, A.G. (2011). Do Red and Green Make Brown?: Perspectives on Plastid Acquisitions within Chromalveolates. Eukaryotic Cell 10, 856–868.

Douglas, A.E. (2003). Coral bleaching––how and why? Marine Pollution Bulletin 46, 385–392.

Douglas, A.E. (2008). Conflict, cheats and the persistence of symbioses. New Phytologist 177, 849–858.

Douglas, A.E. (2014). Symbiosis as a General Principle in Eukaryotic Evolution. Cold Spring Harb Perspect Biol 6, a016113.

Douglas, A., and Smith, D.C. (1984). The green hydra symbiosis. VIII. Mechanisms in symbiont regulation. Proceedings of the Royal Society of London. Series B. Biological Sciences 221, 291–319.

Du, Z.-Y., Zienkiewicz, K., Vande Pol, N., Ostrom, N.E., Benning, C., and Bonito, G.M. (2019). Algal-fungal symbiosis leads to photosynthetic mycelium. ELife 8, e47815.

Dubinsky, Z., and Berman-Frank, I. (2001). Uncoupling primary production from population growth in photosynthesizing organisms in aquatic ecosystems. Aquat. Sci. 63, 4–17.

Dyall, S.D., Brown, M.T., and Johnson, P.J. (2004). Ancient Invasions: From Endosymbionts to Organelles. Science 304, 253–257.

Esteban, G.F., Fenchel, T., and Finlay, B.J. (2010). Mixotrophy in Ciliates. Protist 161, 621–641.

Estrela, S., Kerr, B., and Morris, J.J. (2016). Transitions in individuality through symbiosis. Current Opinion in Microbiology 31, 191–198.

Fabina, N.S., Putnam, H.M., Franklin, E.C., Stat, M., and Gates, R.D. (2013). Symbiotic specificity, association patterns, and function determine community responses to global changes: defining critical research areas for coral-Symbiodinium symbioses. Global Change Biology 19, 3306–3316.

Fenchel, T. (1987). Ecology of Protozoa: The Biology of Free-living Phagotropic Protists (Springer-Verlag).

Ferea, T.L., Botstein, D., Brown, P.O., and Rosenzweig, R.F. (1999). Systematic changes in gene expression patterns following adaptive evolution in yeast. PNAS 96, 9721–9726.

Ferrari, J., Scarborough, C.L., and Godfray, H.C.J. (2007). Genetic variation in the effect of a facultative symbiont on host-plant use by pea aphids. Oecologia 153, 323–329.

Fisher, R.M., Henry, L.M., Cornwallis, C.K., Kiers, E.T., and West, S.A. (2017). The evolution of host-symbiont dependence. Nature Communications 8, ncomms15973.

Focke, G.W. (1836). Ueber einige Organisation-sverhaltnisse bei polygastrischen Infusorien, “Oken. Isis,.”

143

Fokin, S.I. (2004). Bacterial Endocytobionts of Ciliophora and Their Interactions with the Host Cell. In International Review of Cytology, (Academic Press), pp. 181–249.

Foyer, C.H., and Noctor, G. (2003). Redox sensing and signalling associated with reactive oxygen in chloroplasts, peroxisomes and mitochondria. Physiologia Plantarum 119, 355–364.

Frago, E., Dicke, M., and Godfray, H.C.J. (2012). Insect symbionts as hidden players in insect–plant interactions. Trends in Ecology & Evolution 27, 705–711.

Frank, S.A. (1997). Models of Symbiosis. The American Naturalist 150, S80–S99.

Freeman, C.J., Thacker, R.W., Baker, D.M., and Fogel, M.L. (2013). Quality or quantity: is nutrient transfer driven more by symbiont identity and productivity than by symbiont abundance? The ISME Journal 7, 1116–1125.

Fujishima, M. (2009). Endosymbionts in Paramecium (Springer Science & Business Media).

Fujishima, M., and Kodama, Y. (2012). Endosymbionts in Paramecium. European Journal of Protistology 48, 124–137.

Fukatsu, T., Aoki, S., Kurosu, U., and Ishikawa, H. (1994). Phylogeny of Cerataphidini Aphids Revealed by Their Symbiotic Microorganisms and Basic Structure of Their Galls–Implications for Host-Symbiont Coevolution and Evolution of Sterile Soldier Castes. Zoological Science 11, p613-623.

Futuyma, D.J., and Moreno, G. (1988). The Evolution of Ecological Specialization. Annual Review of Ecology and Systematics 19, 207–233.

Garcia, J.R., and Gerardo, N.M. (2014). The symbiont side of symbiosis: do microbes really benefit? Front. Microbiol. 5.

Gargas, A., DePriest, P.T., Grube, M., and Tehler, A. (1995). Multiple origins of lichen symbioses in fungi suggested by SSU rDNA phylogeny. Science 268, 1492–1495.

Gast, R.J., Sanders, R.W., and Caron, D.A. (2009). Ecological strategies of protists and their symbiotic relationships with prokaryotic microbes. Trends Microbiol. 17, 563–569.

Genkai-Kato, M., and Yamamura, N. (1999). Evolution of Mutualistic Symbiosis without Vertical Transmission. Theoretical Population Biology 55, 309–323.

Gilbert, S.F., McDonald, E., Boyle, N., Buttino, N., Gyi, L., Mai, M., Neelakantan, P., and James, R. (2010). Symbiosis as a source of selectable epigenetic variation: taking the heat for the big guy. Philosophical Transactions of the Royal Society B: Biological Sciences 365, 671–678.

Gong, J., Qing, Y., Guo, X., and Warren, A. (2014). “Candidatus Sonnebornia yantaiensis”, a member of candidate division OD1, as intracellular bacteria of the ciliated protist Paramecium bursaria (Ciliophora, Oligohymenophorea). Systematic and Applied Microbiology 37, 35–41.

144

Gornik, S.G., Febrimarsa, Cassin, A.M., MacRae, J.I., Ramaprasad, A., Rchiad, Z., McConville, M.J., Bacic, A., McFadden, G.I., Pain, A., et al. (2015). Endosymbiosis undone by stepwise elimination of the plastid in a parasitic dinoflagellate. PNAS 112, 5767–5772.

Graf, A., Schlereth, A., Stitt, M., and Smith, A.M. (2010). Circadian control of carbohydrate availability for growth in Arabidopsis plants at night. PNAS 107, 9458– 9463.

Gray, M.W., and Doolittle, W.F. (1982). Has the endosymbiont hypothesis been proven? Microbiol Rev 46, 1–42.

Hamada, M., Schröder, K., Bathia, J., Kürn, U., Fraune, S., Khalturina, M., Khalturin, K., Shinzato, C., Satoh, N., and Bosch, T.C. (2018). Metabolic co-dependence drives the evolutionarily ancient Hydra–Chlorella symbiosis. ELife Sciences 7, e35122.

Hardin, G. (1968). The Tragedy of the Commons. Science 162, 1243–1248.

He, M., Wang, J., Fan, X., Liu, X., Shi, W., Huang, N., Zhao, F., and Miao, M. (2019). Genetic basis for the establishment of endosymbiosis in Paramecium. The ISME Journal 13, 1360.

Heath, K.D. (2010). Intergenomic Epistasis and Coevolutionary Constraint in Plants and Rhizobia. Evolution 64, 1446–1458.

Heath, K.D., and Tiffin, P. (2007). Context dependence in the coevolution of plant and rhizobial mutualists. Proceedings of the Royal Society of London B: Biological Sciences 274, 1905–1912.

Heath, K.D., Stock, A.J., and Stinchcombe, J.R. (2010). Mutualism variation in the nodulation response to nitrate. Journal of Evolutionary Biology 23, 2494–2500.

Herre, E.A., Knowlton, N., Mueller, U.G., and Rehner, S.A. (1999). The evolution of mutualisms: exploring the paths between conflict and cooperation. Trends in Ecology & Evolution 14, 49–53.

Higuchi, R., Song, C., Hoshina, R., and Suzaki, T. (2018). Endosymbiosis-related changes in ultrastructure and chemical composition of Chlorella variabilis (Archaeplastida, Chlorophyta) cell wall in Paramecium bursaria (Ciliophora, Oligohymenophorea). European Journal of Protistology 66, 149–155.

Hill, D.J. (2009). Asymmetric Co-evolution in the Lichen Symbiosis Caused by a Limited Capacity for Adaptation in the Photobiont. Bot. Rev. 75, 326–338.

Hoang, K.L., Morran, L.T., and Gerardo, N.M. (2016). Experimental Evolution as an Underutilized Tool for Studying Beneficial Animal–Microbe Interactions. Front. Microbiol. 7. van Hoek, A.H.A.M., van Alen, T.A., Sprakel, V.S.I., Leunissen, J.A.M., Brigge, T., Vogels, G.D., and Hackstein, J.H.P. (2000). Multiple Acquisition of Methanogenic Archaeal Symbionts by Anaerobic Ciliates. Mol Biol Evol 17, 251–258.

145

Holland, J.N., DeAngelis, D.L., and Bronstein, J.L. (2002). Population Dynamics and Mutualism: Functional Responses of Benefits and Costs. The American Naturalist 159, 231–244.

Holland, J.N., DeAngelis, D.L., and Schultz, S.T. (2004). Evolutionary stability of mutualism: interspecific population regulation as an evolutionarily stable strategy. Proceedings of the Royal Society of London. Series B: Biological Sciences 271, 1807– 1814.

Honegger, R. (1991). Functional Aspects of the Lichen Symbiosis. Annual Review of Plant Physiology and Plant Molecular Biology 42, 553–578.

Hoogenboom, M., Beraud, E., and Ferrier-Pagès, C. (2010). Relationship between symbiont density and photosynthetic carbon acquisition in the temperate coral Cladocora caespitosa. Coral Reefs 29, 21–29.

Hopkins, D.P., Cameron, D.D., and Butlin, R.K. (2017). The chemical signatures underlying host plant discrimination by aphids. Scientific Reports 7, 8498.

Hörtnagl, P.H., and Sommaruga, R. (2007). Photo-oxidative stress in symbiotic and aposymbiotic strains of the ciliate Paramecium bursaria. Photochemical & Photobiological Sciences 6, 842.

Hoshina, R., and Fujiwara, Y. (2012). Photobiont Flexibility in Paramecium bursaria: Double and Triple Photobiont Co-Habitation. Advances in Microbiology 2, 227–233.

Hoshina, R., and Imamura, N. (2008). Multiple Origins of the Symbioses in Paramecium bursaria. Protist 159, 53–63.

Hoshina, R., Kamako, S. -i, and Imamura, N. (2004). Phylogenetic Position of Endosymbiotic Green Algae in Paramecium bursaria Ehrenberg from Japan. Plant Biol (Stuttg) 6, 447–453.

Hoshina, R., Kato, Y., Kamako, S., and Imamura, N. (2005). Genetic Evidence of “American” and “European” Type Symbiotic Algae of Paramecium bursaria Ehrenberg. Plant Biol (Stuttg) 7, 526–532.

Hosokawa, T., Kikuchi, Y., Shimada, M., and Fukatsu, T. (2007). Obligate symbiont involved in pest status of host insect. Proceedings of the Royal Society B: Biological Sciences 274, 1979–1984.

Hotopp, J.C.D., Clark, M.E., Oliveira, D.C.S.G., Foster, J.M., Fischer, P., Torres, M.C.M., Giebel, J.D., Kumar, N., Ishmael, N., Wang, S., et al. (2007). Widespread Lateral Gene Transfer from Intracellular Bacteria to Multicellular Eukaryotes. Science 317, 1753–1756.

Howells, E.J., Beltran, V.H., Larsen, N.W., Bay, L.K., Willis, B.L., and van Oppen, M.J.H. (2012). Coral thermal tolerance shaped by local adaptation of photosymbionts. Nature Climate Change 2, 116–120.

146

Hughes, T.P., Baird, A.H., Bellwood, D.R., Card, M., Connolly, S.R., Folke, C., Grosberg, R., Hoegh-Guldberg, O., Jackson, J.B.C., Kleypas, J., et al. (2003). Climate Change, Human Impacts, and the Resilience of Coral Reefs. Science 301, 929–933.

Hughes, T.P., Kerry, J.T., Álvarez-Noriega, M., Álvarez-Romero, J.G., Anderson, K.D., Baird, A.H., Babcock, R.C., Beger, M., Bellwood, D.R., Berkelmans, R., et al. (2017). Global warming and recurrent mass bleaching of corals. Nature 543, 373–377.

Ibarra, R.U., Edwards, J.S., and Palsson, B.O. (2002). Escherichia coli K-12 undergoes adaptive evolution to achieve in silico predicted optimal growth. Nature 420, 186–189.

Ishikawa, M., Yuyama, I., Shimizu, H., Nozawa, M., Ikeo, K., and Gojobori, T. (2016). Different Endosymbiotic Interactions in Two Hydra Species Reflect the Evolutionary History of Endosymbiosis. Genome Biol Evol 8, 2155–2163.

Jaenike, J., Unckless, R., Cockburn, S.N., Boelio, L.M., and Perlman, S.J. (2010). Adaptation via Symbiosis: Recent Spread of a Drosophila Defensive Symbiont. Science 329, 212–215.

Jeon, K.W. (1987). Change of Cellular “Pathogens” into Required Cell Componentsa. Annals of the New York Academy of Sciences 503, 359–371.

Jessup, C.M., Kassen, R., Forde, S.E., Kerr, B., Buckling, A., Rainey, P.B., and Bohannan, B.J.M. (2004). Big questions, small worlds: microbial model systems in ecology. Trends in Ecology & Evolution 19, 189–197.

Jiggins, F.M., and Hurst, G.D.D. (2011). Rapid Insect Evolution by Symbiont Transfer. Science 332, 185–186.

Johnson, M.D. (2011). The acquisition of phototrophy: adaptive strategies of hosting endosymbionts and organelles. Photosynth Res 107, 117–132.

Johnstone, R.A., and Bshary, R. (2002). From parasitism to mutualism: partner control in asymmetric interactions. Ecology Letters 5, 634–639.

Joy, J.B. (2013). Symbiosis catalyses niche expansion and diversification. Proceedings of the Royal Society B: Biological Sciences 280, 20122820.

Kadono, T., Kawano, T., Hosoya, H., and Kosaka, T. (2004). Flow cytometric studies of the host-regulated cell cycle in algae symbiotic with green paramecium. Protoplasma 223, 133–141.

Kaever, A., Lingner, T., Feussner, K., Göbel, C., Feussner, I., and Meinicke, P. (2009). MarVis: a tool for clustering and visualization of metabolic biomarkers. BMC Bioinformatics 10, 92.

Kaltenpoth, M., Roeser-Mueller, K., Koehler, S., Peterson, A., Nechitaylo, T.Y., Stubblefield, J.W., Herzner, G., Seger, J., and Strohm, E. (2014). Partner choice and fidelity stabilize coevolution in a Cretaceous-age defensive symbiosis. PNAS 111, 6359– 6364.

147

Kamako, S.-I., and Imamura, N. (2006). Effect of Japanese Paramecium bursaria Extract on Photosynthetic Carbon Fixation of Symbiotic Algae. Journal of Eukaryotic Microbiology 53, 136–141.

Kamako, S., Hoshina, R., Ueno, S., and Imamura, N. (2005). Establishment of axenic endosymbiotic strains of Japanese Paramecium bursaria and the utilization of carbohydrate and nitrogen compounds by the isolated algae. European Journal of Protistology 41, 193–202.

Kanehisa, M., and Goto, S. (2000). KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 28, 27–30.

Kanehisa, M., Sato, Y., Furumichi, M., Morishima, K., and Tanabe, M. (2019). New approach for understanding genome variations in KEGG. Nucleic Acids Res. 47, D590– D595.

Karakashian, S.J. (1963). Growth of Paramecium bursaria as Influenced by the Presence of Algal Symbionts. Physiological Zoology 36, 52–68.

Karkar, S., Facchinelli, F., Price, D.C., Weber, A.P.M., and Bhattacharya, D. (2015). Metabolic connectivity as a driver of host and endosymbiont integration. PNAS 112, 10208–10215.

Karnkowska, A., Vacek, V., Zubáčová, Z., Treitli, S.C., Petrželková, R., Eme, L., Novák, L., Žárský, V., Barlow, L.D., Herman, E.K., et al. (2016). A Eukaryote without a Mitochondrial Organelle. Current Biology 26, 1274–1284.

Kato, Y., and Imamura, N. (2008a). Effect of calcium ion on uptake of amino acids by symbiotic Chlorella F36-ZK isolated from Japanese Paramecium bursaria. Plant Science 174, 88–96.

Kato, Y., and Imamura, N. (2008b). Effect of sugars on amino acid transport by symbiotic Chlorella. Plant Physiol. Biochem. 46, 911–917.

Kato, Y., Ueno, S., and Imamura, N. (2006). Studies on the nitrogen utilization of endosymbiotic algae isolated from Japanese Paramecium bursaria. Plant Science 170, 481–486.

Kawano, T., Kadono, T., Kosaka, T., and Hosoya, H. (2004). Green Paramecia as an Evolutionary Winner of Oxidative Symbiosis: A Hypothesis and Supportive Data. Zeitschrift Für Naturforschung C 59, 538–542.

Keeling, P.J. (2010). The endosymbiotic origin, diversification and fate of plastids. Philosophical Transactions of the Royal Society B: Biological Sciences 365, 729–748.

Keeling, P.J. (2013). The number, speed, and impact of plastid endosymbioses in eukaryotic evolution. Annu Rev Plant Biol 64, 583–607.

Keeling, P.J., and Archibald, J.M. (2008). Organelle Evolution: What’s in a Name? Current Biology 18, R345–R347.

148

Keeling, P.J., and McCutcheon, J.P. (2017). Endosymbiosis: The feeling is not mutual. J. Theor. Biol. 434, 75–79.

Kessler, E., and Huss, V. a. R. (1990). Biochemical Taxonomy of Symbiotic Chlorella Strains from Paramecium and Acanthocystis*. Botanica Acta 103, 140–142.

Kiers, E.T., and West, S.A. (2015). Evolving new organisms via symbiosis. Science 348, 392–394.

Kiers, E.T., Rousseau, R.A., West, S.A., and Denison, R.F. (2003). Host sanctions and the legume–rhizobium mutualism. Nature 425, 78–81.

Kikuchi, Y., Tada, A., Musolin, D.L., Hari, N., Hosokawa, T., Fujisaki, K., and Fukatsu, T. (2016). Collapse of Insect Gut Symbiosis under Simulated Climate Change. MBio 7, e01578-16.

King, K.C., Brockhurst, M.A., Vasieva, O., Paterson, S., Betts, A., Ford, S.A., Frost, C.L., Horsburgh, M.J., Haldenby, S., and Hurst, G.D. (2016). Rapid evolution of microbe-mediated protection against pathogens in a worm host. The ISME Journal 10, 1915–1924.

Kiseleva, A.A., Tarachovskaya, E.R., and Shishova, M.F. (2012). Biosynthesis of phytohormones in algae. Russ J Plant Physiol 59, 595–610.

Klyachko‐Gurvich, G.L., Tsoglin, L.N., Doucha, J., Kopetskii, J., Shebalina (ryabykh), I.B., and Semenenko, V.E. (1999). Desaturation of fatty acids as an adaptive response to shifts in light intensity. Physiologia Plantarum 107, 240–249.

Knight, R., Ley, R.E., Raes, J., and Grice, E.A. (2019). Expanding the scope and scale of microbiome research. Genome Biology 20, 191.

Koch, A.M., Antunes, P.M., Maherali, H., Hart, M.M., and Klironomos, J.N. (2017). Evolutionary asymmetry in the arbuscular mycorrhizal symbiosis: conservatism in fungal morphology does not predict host plant growth. New Phytologist 214, 1330–1337.

Kodama, Y., and Fujishima, M. (2008). Cycloheximide Induces Synchronous Swelling of Perialgal Vacuoles Enclosing Symbiotic Chlorella vulgaris and Digestion of the Algae in the Ciliate Paramecium bursaria. Protist 159, 483–494.

Kodama, Y., and Fujishima, M. (2011). Four important cytological events needed to establishG endosymbiosis of symbiotic Chlorella sp. to the alga-free Paramecium bursaria. Jpn. J. Protozool. Vol 44, 1.

Kodama, Y., and Fujishima, M. (2012). Cell division and density of symbiotic Chlorella variabilis of the ciliate Paramecium bursaria is controlled by the host’s nutritional conditions during early infection process. Environmental Microbiology 14, 2800–2811.

Kodama, Y., and Fujishima, M. (2014). Symbiotic Chlorella variabilis incubated under constant dark conditions for 24 hours loses the ability to avoid digestion by host lysosomal enzymes in digestive vacuoles of host ciliate Paramecium bursaria. FEMS Microbiol Ecol 90, 946–955.

149

Kodama, Y., Suzuki, H., Dohra, H., Sugii, M., Kitazume, T., Yamaguchi, K., Shigenobu, S., and Fujishima, M. (2014). Comparison of gene expression of Paramecium bursaria with and without Chlorella variabilissymbionts. BMC Genomics 15, 183.

Kodama, Y., Nagase, M., and Takahama, A. (2016). Symbiotic Chlorella variabilis strain, 1 N, can influence the digestive process in the host Paramecium bursaria during early infection. Symbiosis 1–9.

Koga, R., and Moran, N.A. (2014). Swapping symbionts in spittlebugs: evolutionary replacement of a reduced genome symbiont. The ISME Journal 8, 1237–1246.

Koga, R., Tsuchida, T., Sakurai, M., and Fukatsu, T. (2007). Selective elimination of aphid endosymbionts: effects of antibiotic dose and host genotype, and fitness consequences. FEMS Microbiol Ecol 60, 229–239.

Kondo, N., Shimada, M., and Fukatsu, T. (2005). Infection density of Wolbachia endosymbiont affected by co-infection and host genotype. Biology Letters 1, 488–491.

Kour, D., Rana, K.L., Yadav, N., Yadav, A.N., Kumar, A., Meena, V.S., Singh, B., Chauhan, V.S., Dhaliwal, H.S., and Saxena, A.K. (2019). Rhizospheric Microbiomes: Biodiversity, Mechanisms of Plant Growth Promotion, and Biotechnological Applications for Sustainable Agriculture. In Plant Growth Promoting Rhizobacteria for Agricultural Sustainability : From Theory to Practices, A. Kumar, and V.S. Meena, eds. (Singapore: Springer Singapore), pp. 19–65.

Kreutz, M., Stoeck, T., and Foissner, W. (2012). Morphological and Molecular Characterization of Paramecium (Viridoparamecium nov. subgen.) chlorelligerum Kahl (Ciliophora). Journal of Eukaryotic Microbiology 59, 548–563.

Kropp, B.R., and Trappe, J.M. (1982). Ectomycorrhizal Fungi of Tsuga Heterophylla. Mycologia 74, 479–488.

Lane, C.E., and Archibald, J.M. (2008). The eukaryotic tree of life: endosymbiosis takes its TOL. Trends in Ecology & Evolution 23, 268–275.

Lanzoni, O., Fokin, S.I., Lebedeva, N., Migunova, A., Petroni, G., and Potekhin, A. (2016). Rare Freshwater Ciliate Paramecium chlorelligerum Kahl, 1935 and Its Macronuclear Symbiotic Bacterium “Candidatus Holospora parva.” PLOS ONE 11, e0167928.

Larkum, A.W.D., Lockhart, P.J., and Howe, C.J. (2007). Shopping for plastids. Trends in Plant Science 12, 189–195.

Law, R., and Dieckmann, U. (1998). Symbiosis through exploitation and the merger of lineages in evolution. Proceedings of the Royal Society of London B: Biological Sciences 265, 1245–1253.

Lefèvre, C., Charles, H., Vallier, A., Delobel, B., Farrell, B., and Heddi, A. (2004). Endosymbiont Phylogenesis in the Dryophthoridae Weevils: Evidence for Bacterial Replacement. Mol Biol Evol 21, 965–973.

150

Lesser, M.P. (2006). OXIDATIVE STRESS IN MARINE ENVIRONMENTS: Biochemistry and Physiological Ecology. Annual Review of Physiology 68, 253–278.

Lesser, M.P. (2011). Coral Bleaching: Causes and Mechanisms. In Coral Reefs: An Ecosystem in Transition, Z. Dubinsky, and N. Stambler, eds. (Dordrecht: Springer Netherlands), pp. 405–419.

Lewis, N.E., Hixson, K.K., Conrad, T.M., Lerman, J.A., Charusanti, P., Polpitiya, A.D., Adkins, J.N., Schramm, G., Purvine, S.O., Lopez-Ferrer, D., et al. (2010). Omic data from evolved E. coli are consistent with computed optimal growth from genome-scale models. Molecular Systems Biology 6, 390.

Lohse, K., Gutierrez, A., Kaltz, O., and Koella, J. (2006). Experimental evolution of resistance in paramecium caudatum against the bacterial parasite holospora undulata. Evolution 60, 1177–1186.

López-García, P., Eme, L., and Moreira, D. (2017). Symbiosis in eukaryotic evolution. Journal of Theoretical Biology 434, 20–33.

Lowe, C.D., Minter, E.J., Cameron, D.D., and Brockhurst, M.A. (2016). Shining a Light on Exploitative Host Control in a Photosynthetic Endosymbiosis. Current Biology 26, 207–211.

Loya, Y., Sakai, K., Yamazato, K., Nakano, Y., Sambali, H., and Woesik, R. van (2001). Coral bleaching: the winners and the losers. Ecology Letters 4, 122–131.

Lu, Y., Tarkowská, D., Turečková, V., Luo, T., Xin, Y., Li, J., Wang, Q., Jiao, N., Strnad, M., and Xu, J. (2014). Antagonistic roles of abscisic acid and cytokinin during response to nitrogen depletion in oleaginous microalga Nannochloropsis oceanica expand the evolutionary breadth of phytohormone function. The Plant Journal 80, 52–68.

Maharjan, R.P., Seeto, S., and Ferenci, T. (2007). Divergence and Redundancy of Transport and Metabolic Rate-Yield Strategies in a Single Escherichia coli Population. Journal of Bacteriology 189, 2350–2358.

Mallick, N. (2004). Copper-induced oxidative stress in the chlorophycean microalga Chlorella vulgaris: response of the antioxidant system. Journal of Plant Physiology 161, 591–597.

Manoharan, K., Lee, T.K., Cha, J.M., Kim, J.H., Lee, W.S., Chang, M., Park, C.W., and Cho, J.H. (1999). Acclimation of Prorocentrum Minimum (dinophyceae) to Prolonged Darkness by Use of an Alternative Carbon Source from Triacylglycerides and Galactolipids. Journal of Phycology 35, 287–292.

Marin, B., M. Nowack, E.C., and Melkonian, M. (2005). A Plastid in the Making: Evidence for a Second Primary Endosymbiosis. Protist 156, 425–432.

Martin, W.F., Garg, S., and Zimorski, V. (2015). Endosymbiotic theories for eukaryote origin. Philosophical Transactions of the Royal Society B: Biological Sciences 370, 20140330.

151

Masson-Boivin, C., Giraud, E., Perret, X., and Batut, J. (2009). Establishing nitrogen- fixing symbiosis with legumes: how many rhizobium recipes? Trends in Microbiology 17, 458–466.

Matsuura, Y., Moriyama, M., Łukasik, P., Vanderpool, D., Tanahashi, M., Meng, X.-Y., McCutcheon, J.P., and Fukatsu, T. (2018). Recurrent symbiont recruitment from fungal parasites in cicadas. PNAS 115, E5970–E5979.

Matthews, J.L., Oakley, C.A., Lutz, A., Hillyer, K.E., Roessner, U., Grossman, A.R., Weis, V.M., and Davy, S.K. (2018). Partner switching and metabolic flux in a model cnidarian–dinoflagellate symbiosis. Proceedings of the Royal Society B: Biological Sciences 285, 20182336.

Maxwell, K., and Johnson, G.N. (2000). Chlorophyll fluorescence—a practical guide. J Exp Bot 51, 659–668.

McCutcheon, J.P., and Moran, N.A. (2012). Extreme genome reduction in symbiotic bacteria. Nature Reviews Microbiology 10, 13–26.

McCutcheon, J.P., McDonald, B.R., and Moran, N.A. (2009). Origin of an Alternative Genetic Code in the Extremely Small and GC–Rich Genome of a Bacterial Symbiont. PLOS Genetics 5, e1000565.

McFall-Ngai, M.J., and Ruby, E.G. (1991). Symbiont recognition and subsequent morphogenesis as early events in an animal-bacterial mutualism. Science 254, 1491– 1494.

McFall-Ngai, M.J., and Ruby, E.G. (1998). Sepiolids and Vibrios: When First They Meet. BioScience 48, 257–265.

McGraw, E.A., Merritt, D.J., Droller, J.N., and O’Neill, S.L. (2002). Wolbachia density and virulence attenuation after transfer into a novel host. PNAS 99, 2918–2923.

McKew, B.A., Davey, P., Finch, S.J., Hopkins, J., Lefebvre, S.C., Metodiev, M.V., Oxborough, K., Raines, C.A., Lawson, T., and Geider, R.J. (2013). The trade-off between the light-harvesting and photoprotective functions of fucoxanthin-chlorophyll proteins dominates light acclimation in Emiliania huxleyi (clone CCMP 1516). New Phytologist 200, 74–85.

Mereschkowsky, C. (1905). Über Natur und Ursprung der Chromatophoren im Pflanzenreiche. Biologisches Centralblatt 25/18: 38–604, ed. J. Rosenthal.

Merzlyak, M.N., Chivkunova, O.B., Gorelova, O.A., Reshetnikova, I.V., Solovchenko, A.E., Khozin‐Goldberg, I., and Cohen, Z. (2007). Effect of Nitrogen Starvation on Optical Properties, Pigments, and Arachidonic Acid Content of the Unicellular Green Alga Parietochloris Incisa (trebouxiophyceae, Chlorophyta)1. Journal of Phycology 43, 833–843.

Mews, L.K. (1980). The green hydra symbiosis. III. The biotrophic transport of carbohydrate from alga to animal. Proceedings of the Royal Society of London. Series B. Biological Sciences 209, 377–401.

152

Minaeva, E., and Ermilova, E. (2017). Responses triggered in chloroplast of Chlorella variabilis NC64A by long-term association with Paramecium bursaria. Protoplasma 254, 1769–1776.

Minter, E.J.A., Lowe, C.D., Sørensen, M.E.S., Wood, A.J., Cameron, D.D., and Brockhurst, M.A. (2018). Variation and asymmetry in host-symbiont dependence in a microbial symbiosis. BMC Evolutionary Biology 18, 108.

Miwa, I., Fujimori, N., and Tanaka, M. (1996). Effects of symbiotic Chlorella on the period length and the phase shift of circadian rhythms in Paramecium bursaria. European Journal of Protistology 32, 102–107.

Molero, G., Aranjuelo, I., Teixidor, P., Araus, J.L., and Nogués, S. (2011). Measurement of 13C and 15N isotope labeling by gas chromatography/combustion/isotope ratio mass spectrometry to study amino acid fluxes in a plant–microbe symbiotic association. Rapid Communications in Mass Spectrometry 25, 599–607.

Molina, R., and Trappe, J.M. (1982). Patterns of Ectomycorrhizal Host Specificity and Potential among Pacific Northwest Conifers and Fungi. For Sci 28, 423–458.

Monika, Devi, S., Arya, S.S., Kumar, N., and Kumar, S. (2019). Mycorrhizal Fungi: Potential Candidate for Sustainable Agriculture. In Mycorrhizosphere and Pedogenesis, A. Varma, and D.K. Choudhary, eds. (Singapore: Springer Singapore), pp. 339–353.

Moran, N.A. (2007). Symbiosis as an adaptive process and source of phenotypic complexity. PNAS 104, 8627–8633.

Moran, N.A., and Wernegreen, J.J. (2000). Lifestyle evolution in symbiotic bacteria: insights from genomics. Trends in Ecology & Evolution 15, 321–326.

Moran, N.A., Plague, G.R., Sandström, J.P., and Wilcox, J.L. (2003). A genomic perspective on nutrient provisioning by bacterial symbionts of insects. PNAS 100, 14543–14548.

Muggia, L., Nelson, P., Wheeler, T., Yakovchenko, L.S., Tønsberg, T., and Spribille, T. (2011). Convergent evolution of a symbiotic duet: The case of the lichen genus Polychidium (Peltigerales, Ascomycota). American Journal of Botany 98, 1647–1656.

Munkacsi, A.B., Pan J. J., Villesen P., Mueller U. G., Blackwell M., and McLaughlin D. J. (2004). Convergent coevolution in the domestication of coral mushrooms by fungus– growing ants. Proceedings of the Royal Society of London. Series B: Biological Sciences 271, 1777–1782.

Murata, N., Takahashi, S., Nishiyama, Y., and Allakhverdiev, S.I. (2007). Photoinhibition of photosystem II under environmental stress. Biochimica et Biophysica Acta (BBA) - Bioenergetics 1767, 414–421.

Muscatine, L., and Porter, J.W. (1977). Reef Corals: Mutualistic Symbioses Adapted to Nutrient-Poor Environments. BioScience 27, 454–460.

153

Nakajima, T., Sano, A., and Matsuoka, H. (2009). Auto-/heterotrophic endosymbiosis evolves in a mature stage of ecosystem development in a microcosm composed of an alga, a bacterium and a ciliate. Biosystems 96, 127–135.

Nakajima, T., Fujikawa, Y., Matsubara, T., Karita, M., and Sano, A. (2015). Differentiation of a free-living alga into forms with ecto- and endosymbiotic associations with heterotrophic organisms in a 5-year microcosm culture. Biosystems 131, 9–21.

Nakayama, S., Parratt, S.R., Hutchence, K.J., Lewis, Z., Price, T. a. R., and Hurst, G.D.D. (2015). Can maternally inherited endosymbionts adapt to a novel host? Direct costs of Spiroplasma infection, but not vertical transmission efficiency, evolve rapidly after horizontal transfer into D. melanogaster. Heredity 114, 539–543.

Neill, S., Desikan, R., and Hancock, J. (2002). Hydrogen peroxide signalling. Current Opinion in Plant Biology 5, 388–395.

Nobre, T. (2019). Symbiosis in Sustainable Agriculture: Can Olive Fruit Fly Bacterial Microbiome Be Useful in Pest Management? Microorganisms 7, 238.

Nowack, E.C. (2014). Paulinella chromatophora - rethinking the transition from endosymbiont to organelle. Acta Societatis Botanicorum Poloniae 83.

Nowack, E.C., and Melkonian, M. (2010). Endosymbiotic associations within protists. Philosophical Transactions of the Royal Society B: Biological Sciences 365, 699–712.

Nowack, E.C.M., Melkonian, M., and Glöckner, G. (2008). Chromatophore Genome Sequence of Paulinella Sheds Light on Acquisition of Photosynthesis by Eukaryotes. Current Biology 18, 410–418.

Nussbaumer, A.D., Fisher, C.R., and Bright, M. (2006). Horizontal endosymbiont transmission in hydrothermal vent tubeworms. Nature 441, 345–348.

Ohkawa, H., Hashimoto, N., Furukawa, S., Kadono, T., and Kawano, T. (2011). Forced symbiosis between Synechocystis spp. PCC 6803 and apo-symbiotic Paramecium bursaria as an experimental model for evolutionary emergence of primitive photosynthetic eukaryotes. Plant Signaling & Behavior 6, 773–776.

O’Neill, S.L. (2018). The Use of Wolbachia by the World Mosquito Program to Interrupt Transmission of Aedes aegypti Transmitted Viruses. Adv. Exp. Med. Biol. 1062, 355– 360. van Oppen, M.J.H., Oliver, J.K., Putnam, H.M., and Gates, R.D. (2015). Building coral reef resilience through assisted evolution. PNAS 112, 2307–2313. van Oppen, M.J.H., Gates, R.D., Blackall, L.L., Cantin, N., Chakravarti, L.J., Chan, W.Y., Cormick, C., Crean, A., Damjanovic, K., Epstein, H., et al. (2017). Shifting paradigms in restoration of the world’s coral reefs. Global Change Biology 23, 3437– 3448.

Ortmayr, K., Causon, T.J., Hann, S., and Koellensperger, G. (2016). Increasing selectivity and coverage in LC-MS based metabolome analysis. TrAC Trends in Analytical Chemistry 82, 358–366.

154

Overy, S.A., Walker, H.J., Malone, S., Howard, T.P., Baxter, C.J., Sweetlove, L.J., Hill, S.A., and Quick, W.P. (2005). Application of metabolite profiling to the identification of traits in a population of tomato introgression lines. J Exp Bot 56, 287–296.

Oxborough, K., Moore, C.M., Suggett, D.J., Lawson, T., Chan, H.G., and Geider, R.J. (2012). Direct estimation of functional PSII reaction center concentration and PSII electron flux on a volume basis: a new approach to the analysis of Fast Repetition Rate fluorometry (FRRf) data. Limnology and Oceanography: Methods 10, 142–154.

Padfield, D., Yvon‐Durocher, G., Buckling, A., Jennings, S., and Yvon‐Durocher, G. (2016). Rapid evolution of metabolic traits explains thermal adaptation in phytoplankton. Ecology Letters 19, 133–142.

Parfrey, L.W., Lahr, D.J.G., Knoll, A.H., and Katz, L.A. (2011). Estimating the timing of early eukaryotic diversification with multigene molecular clocks. PNAS 108, 13624– 13629.

Patron, N.J., Waller, R.F., and Keeling, P.J. (2006). A tertiary plastid uses genes from two endosymbionts. J. Mol. Biol. 357, 1373–1382.

Perez, S., and Weis, V. (2006). Nitric oxide and cnidarian bleaching: an eviction notice mediates breakdown of a symbiosis. Journal of Experimental Biology 209, 2804–2810.

Peters, E. (1996). Prolonged darkness and diatom mortality: II. Marine temperate species. Journal of Experimental Marine Biology and Ecology 207, 43–58.

Pfeiffer, T., Schuster, S., and Bonhoeffer, S. (2001). Cooperation and Competition in the Evolution of ATP-Producing Pathways. Science 292, 504–507.

Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D., and R core Team (2019) (2019). nlme: Linear and Nonlinear Mixed Effects Models.

Poretsky, R., Rodriguez-R, L.M., Luo, C., Tsementzi, D., and Konstantinidis, K.T. (2014). Strengths and Limitations of 16S rRNA Gene Amplicon Sequencing in Revealing Temporal Microbial Community Dynamics. PLOS ONE 9, e93827.

Powell, J.R., and Rillig, M.C. (2018). Biodiversity of arbuscular mycorrhizal fungi and ecosystem function. New Phytol.

Preer, J.R. (2000). Epigenetic Mechanisms Affecting Macronuclear Development in Paramecium and Tetrahymena. Journal of Eukaryotic Microbiology 47, 515–524.

Quispe, C.F., Sonderman, O., Khasin, M., Riekhof, W.R., Van Etten, J.L., and Nickerson, K.W. (2016). Comparative genomics, transcriptomics, and physiology distinguish symbiotic from free-living Chlorella strains. Algal Research 18, 332–340.

R Core Team (2018). R: A Language and Environment for Statistical Computing.

Raina, J.-B., Eme, L., Pollock, F.J., Spang, A., Archibald, J.M., and Williams, T.A. (2018). Symbiosis in the microbial world: from ecology to genome evolution. Biology Open 7, bio032524.

155

Ral, J.-P., Colleoni, C., Wattebled, F., Dauvillée, D., Nempont, C., Deschamps, P., Li, Z., Morell, M.K., Chibbar, R., Purton, S., et al. (2006). Circadian Clock Regulation of Starch Metabolism Establishes GBSSI as a Major Contributor to Amylopectin Synthesis in Chlamydomonas reinhardtii. Plant Physiology 142, 305–317.

Ramsey, J.S., MacDonald, S.J., Jander, G., Nakabachi, A., Thomas, G.H., and Douglas, A.E. (2010). Genomic evidence for complementary purine metabolism in the pea aphid, Acyrthosiphon pisum, and its symbiotic bacterium Buchnera aphidicola. Insect Molecular Biology 19, 241–248.

Rankin, D.J., Bargum, K., and Kokko, H. (2007). The tragedy of the commons in evolutionary biology. Trends in Ecology & Evolution 22, 643–651.

Ratcliffe, R.G., and Shachar‐Hill, Y. (2006). Measuring multiple fluxes through plant metabolic networks. The Plant Journal 45, 490–511.

Rees, T.A., and Hill, S.A. (1994). Metabolic control analysis of plant metabolism. Plant, Cell & Environment 17, 587–599.

Reisser, W. (1976). [The metabolic interactions between Paramecium bursaria Ehrbg. and Chlorella spec. in the Paramecium bursaria-symbiosis. II. Symbiosis-specific properties of the physiology and the cytology of the symbiotic unit and their regulation (author’s transl)]. Arch Microbiol 111, 161–170.

Reisser, W., Burbank, D., Meints, R., Becker, B., and Vanetten, J. (1991). Viruses Distinguish Symbiotic Chlorella Spp of Paramecium-Bursaria. Endocytobiosis Cell Res. 7, 245–251.

Reyes-Prieto, A., Moustafa, A., and Bhattacharya, D. (2008). Multiple Genes of Apparent Algal Origin Suggest Ciliates May Once Have Been Photosynthetic. Current Biology 18, 956–962.

Rigano, C., Di Martino Rigano, V., Vona, V., and Fuggi, A. (1981). Nitrate reductase and glutamine synthetase activities, nitrate and ammonia assimilation, in the unicellular alga Cyanidium caldarium. Arch. Microbiol. 129, 110–114.

Rivera, M.C., and Lake, J.A. (2004). The ring of life provides evidence for a genome fusion origin of eukaryotes. Nature 431, 152–155.

Rocap, G., Larimer, F.W., Lamerdin, J., Malfatti, S., Chain, P., Ahlgren, N.A., Arellano, A., Coleman, M., Hauser, L., Hess, W.R., et al. (2003). Genome divergence in two Prochlorococcus ecotypes reflects oceanic niche differentiation. Nature 424, 1042–1047.

Rodríguez, F., Derelle, E., Guillou, L., Gall, F.L., Vaulot, D., and Moreau, H. (2005). Ecotype diversity in the marine picoeukaryote Ostreococcus (Chlorophyta, Prasinophyceae). Environmental Microbiology 7, 853–859.

Rodriguez-Garcia, I., and Guil-Guerrero, J.L. (2008). Evaluation of the antioxidant activity of three microalgal species for use as dietary supplements and in the preservation of foods. Food Chemistry 108, 1023–1026.

156

Rolshausen, G., Grande, F.D., Sadowska‐Deś, A.D., Otte, J., and Schmitt, I. (2018). Quantifying the climatic niche of symbiont partners in a lichen symbiosis indicates mutualist-mediated niche expansions. Ecography 41, 1380–1392.

Rowan, R. (2004). Thermal adaptation in reef coral symbionts. Nature 430, 742–742.

Russell, J.A., and Moran, N.A. (2005). Horizontal Transfer of Bacterial Symbionts: Heritability and Fitness Effects in a Novel Aphid Host. Appl. Environ. Microbiol. 71, 7987–7994.

Sabater-Muñoz, B., Toft, C., Alvarez-Ponce, D., and Fares, M.A. (2017). Chance and necessity in the genome evolution of endosymbiotic bacteria of insects. ISME J.

Sachs, J.L., and Simms, E.L. (2006). Pathways to mutualism breakdown. Trends in Ecology & Evolution 21, 585–592.

Sachs, J.L., and Wilcox, T.P. (2006). A shift to parasitism in the jellyfish symbiont Symbiodinium microadriaticum. Proceedings of the Royal Society B: Biological Sciences 273, 425–429.

Sachs, J.L., Mueller, U.G., Wilcox, T.P., and Bull, J.J. (2004). The Evolution of Cooperation. The Quarterly Review of Biology 79, 135–160.

Sachs, J.L., Skophammer, R.G., and Regus, J.U. (2011). Evolutionary transitions in bacterial symbiosis. PNAS 108, 10800–10807.

Safi, C., Zebib, B., Merah, O., Pontalier, P.-Y., and Vaca-Garcia, C. (2014). Morphology, composition, production, processing and applications of Chlorella vulgaris: A review. Renewable and Sustainable Energy Reviews 35, 265–278.

Sagan, L. (1967). On the origin of mitosing cells. J. Theor. Biol. 14, 255–274.

Salvaudon, L., Héraudet, V., Shykoff, J.A., and Koella, J. (2005). Parasite-host fitness trade-offs change with parasite identity: genotype-specific interactions in a plant- pathogen system. Evolution 59, 2518–2524.

Sandström, J.P., Russell, J.A., White, J.P., and Moran, N.A. (2001). Independent origins and horizontal transfer of bacterial symbionts of aphids. Molecular Ecology 10, 217–228.

Schatz, G., and Mason, T.L. (1974). The Biosynthesis of Mitochondrial Proteins. Annual Review of Biochemistry 43, 51–87.

Schneider, C.A., Rasband, W.S., and Eliceiri, K.W. (2012). NIH Image to ImageJ: 25 years of image analysis.

Schüßler, A., and Schnepf, E. (1992). Photosynthesis dependent acidification of perialgal vacuoles in theParamedum bursaria/Chlorella symbiosis: Visualization by monensin. Protoplasma 166, 218–222.

Schwarz, Zs., and Kössel, H. (1980). The primary structure of 16S rDNA from Zea mays chloroplast is homologous to E. coli 16S rRNA. Nature 283, 739–742.

157

Shah, N., and Syrett, P.J. (1984). The uptake of guanine and hypoxanthine by marine microalgae. Journal of the Marine Biological Association of the United Kingdom 64, 545–556.

Shapiro, J.W., and Turner, P.E. (2018). Evolution of mutualism from parasitism in experimental virus populations. Evolution 72, 707–712.

Shibata, A., Takahashi, F., Kasahara, M., and Imamura, N. (2016). Induction of Maltose Release by Light in the Endosymbiont Chlorella variabilis of Paramecium bursaria. Protist.

Shih, J.L., Selph, K.E., Wall, C.B., Wallsgrove, N.J., Lesser, M.P., and Popp, B.N. (2019). Trophic Ecology of the Tropical Pacific Sponge Mycale grandis Inferred from Amino Acid Compound-Specific Isotopic Analyses. Microb Ecol.

Shiu, C.-T., and Lee, T.-M. (2005). Ultraviolet-B-induced oxidative stress and responses of the ascorbate–glutathione cycle in a marine macroalga Ulva fasciata. J Exp Bot 56, 2851–2865.

Siegel, R.W. (1960). Hereditary endosymbiosis in Paramecium bursaria. Experimental Cell Research 19, 239–252.

Silverstein, R.N., Correa, A.M.S., and Baker, A.C. (2012). Specificity is rarely absolute in coral–algal symbiosis: implications for coral response to climate change. Proceedings of the Royal Society B: Biological Sciences 279, 2609–2618.

Singh, D.P., Saudemont, B., Guglielmi, G., Arnaiz, O., Goût, J.-F., Prajer, M., Potekhin, A., Przybòs, E., Aubusson-Fleury, A., Bhullar, S., et al. (2014). Genome-defence small RNAs exapted for epigenetic mating-type inheritance. Nature 509, 447–452.

Smith, G.J., and Muscatine, L. (1999). Cell cycle of symbiotic dinoflagellates: variation in G1 phase-duration with anemone nutritional status and macronutrient supply in the Aiptasia pulchella–Symbiodinium pulchrorum symbiosis. Marine Biology 134, 405–418.

Smith, C.A., O’Maille, G., Want, E.J., Qin, C., Trauger, S.A., Brandon, T.R., Custodio, D.E., Abagyan, R., and Siuzdak, G. (2005). METLIN: a metabolite mass spectral database. Ther Drug Monit 27, 747–751.

Smith, C.A., Want, E.J., O’Maille, G., Abagyan, R., and Siuzdak, G. (2006). XCMS: Processing Mass Spectrometry Data for Metabolite Profiling Using Nonlinear Peak Alignment, Matching, and Identification. Anal. Chem. 78, 779–787.

Soldo, A.T., Godoy, G.A., and Larin, F. (1978). Purine-Excretory Nature of Refractile Bodies in the Marine Ciliate Parauronema acutum*. The Journal of Protozoology 25, 416–418.

Sørensen, M.E., Wood, A.J., Minter, E.J., Lowe, C.D., Cameron, D.D., and Brockhurst, M.A. (2020). Comparison of independent evolutionary origins reveals both convergence and divergence in the metabolic mechanisms of symbiosis. Current Biology.

158

Sørensen, M.E.S., Lowe, C.D., Minter, E.J.A., Wood, A.J., Cameron, D.D., and Brockhurst, M.A. (2019). The role of exploitation in the establishment of mutualistic microbial symbioses. FEMS Microbiol Lett 366.

Spang, A., Saw, J.H., Jørgensen, S.L., Zaremba-Niedzwiedzka, K., Martijn, J., Lind, A.E., van Eijk, R., Schleper, C., Guy, L., and Ettema, T.J.G. (2015). Complex archaea that bridge the gap between prokaryotes and eukaryotes. Nature 521, 173–179.

Sprent, J.I., Sutherland, J.M., and Faria, S.M.D. (1987). Some aspects of the biology of nitrogen-fixing organisms. Phil. Trans. R. Soc. Lond. B 317, 111–129.

Stanier, R.Y. (1970). Some aspects of the biology of cells and their possible evolutionary significance. In Symp Soc Gen Microbiol, pp. 1–38.

Stanley, G.D., and Lipps, J.H. (2011). Photosymbiosis: The Driving Force for Reef Success and Failure. The Paleontological Society Papers 17, 33–59.

Stat, M., Carter, D., and Hoegh-Guldberg, O. (2006). The evolutionary history of Symbiodinium and scleractinian hosts—Symbiosis, diversity, and the effect of climate change. Perspectives in Plant Ecology, Evolution and Systematics 8, 23–43.

Stein, J.R. (1979). (ED.) Handbook of Phycological Methods: Culture Methods and Growth Measurements (Cambridge University Press).

Stiller, J.W., Schreiber, J., Yue, J., Guo, H., Ding, Q., and Huang, J. (2014). The evolution of photosynthesis in chromist algae through serial endosymbioses. Nature Communications 5, 5764.

Stoecker, D.K., Johnson, M.D., Vargas, C. de, and Not, F. (2009). Acquired phototrophy in aquatic protists. Aquat Microb Ecol 57, 279–310.

Storey, J.D., and Tibshirani, R. (2003). Statistical significance for genomewide studies. PNAS 100, 9440–9445.

Straub, C.S., Ives, A.R., and Gratton, C. (2011). Evidence for a Trade-Off between Host- Range Breadth and Host-Use Efficiency in Aphid Parasitoids. The American Naturalist 177, 389–395.

Sudakaran, S., Kost, C., and Kaltenpoth, M. (2017). Symbiont Acquisition and Replacement as a Source of Ecological Innovation. Trends in Microbiology 25, 375–390.

Sugio, A., Dubreuil, G., Giron, D., and Simon, J.-C. (2015). Plant–insect interactions under bacterial influence: ecological implications and underlying mechanisms. J Exp Bot 66, 467–478.

Sully, S., Burkepile, D.E., Donovan, M.K., Hodgson, G., and Woesik, R. van (2019). A global analysis of coral bleaching over the past two decades. Nat Commun 10, 1–5.

Summerer, M., Sonntag, B., and Sommaruga, R. (2007). An experimental test of the symbiosis specificity between the ciliate Paramecium bursaria and strains of the unicellular green alga Chlorella. Environmental Microbiology 9, 2117–2122.

159

Summerer, M., Sonntag, B., and Sommaruga, R. (2008). Ciliate-Symbiont Specificity of Freshwater Endosymbiotic Chlorella (trebouxiophyceae, Chlorophyta)1. Journal of Phycology 44, 77–84.

Summerer, M., Sonntag, B., Hörtnagl, P., and Sommaruga, R. (2009). Symbiotic Ciliates Receive Protection Against UV Damage from their Algae: A Test with Paramecium bursaria and Chlorella. Protist 160, 233–243.

Szathmáry, E., and Smith, J.M. (1995). The major evolutionary transitions. Nature 374, 227–232.

Takahashi, T., Shirai, Y., Kosaka, T., and Hosoya, H. (2007). Arrest of Cytoplasmic Streaming Induces Algal Proliferation in Green Paramecia. PLOS ONE 2, e1352.

Takeda, H. (1988). Classification of Chlorella strains by cell wall sugar composition. Phytochemistry 27, 3823–3826.

Tarakhovskaya, E.R., Maslov, Yu.I., and Shishova, M.F. (2007). Phytohormones in algae. Russ J Plant Physiol 54, 163–170.

Tautenhahn, R., Böttcher, C., and Neumann, S. (2008). Highly sensitive feature detection for high resolution LC/MS. BMC Bioinformatics 9, 504.

Tebo, B.M., Scott Linthicum, D., and Nealson, K.H. (1979). Luminous bacteria and light emitting fish: Ultrastructure of the symbiosis. Biosystems 11, 269–280.

Thompson, G.A. (1996). Lipids and membrane function in green algae. Biochimica et Biophysica Acta (BBA) - Lipids and Lipid Metabolism 1302, 17–45.

Thompson, J.N. (2005). The Geographic Mosaic of Coevolution (University of Chicago Press).

Titlyanov, E., Titlyanova, T., Leletkin, V., Tsukahara, J., van Woesik, R., and Yamazato, K. (1996). Degradation of zooxanthellae and regulation of their density in hermatypic corals. Marine Ecology Progress Series 139, 167–178.

Tonooka, Y., and Watanabe, T. (2002). A natural strain of Paramecium bursaria lacking symbiotic algae. European Journal of Protistology 38, 55–58.

Touw, W.G., Bayjanov, J.R., Overmars, L., Backus, L., Boekhorst, J., Wels, M., and van Hijum, S.A.F.T. (2013). Data mining in the Life Sciences with Random Forest: a walk in the park or lost in the jungle? Brief. Bioinformatics 14, 315–326.

Tremblay, P., Fine, M., Maguer, J.F., Grover, R., and Ferrier-Pagès, C. (2013). Photosynthate translocation increases in response to low seawater pH in a coral– dinoflagellate symbiosis. Biogeosciences 10, 3997–4007.

Treseder, K.K. (2004). A meta-analysis of mycorrhizal responses to nitrogen, phosphorus, and atmospheric CO2 in field studies. New Phytologist 164, 347–355.

160

Tso, G.H.W., Reales-Calderon, J.A., Tan, A.S.M., Sem, X., Le, G.T.T., Tan, T.G., Lai, G.C., Srinivasan, K.G., Yurieva, M., Liao, W., et al. (2018). Experimental evolution of a fungal pathogen into a gut symbiont. Science 362, 589–595.

Tsuchida, T., Koga, R., Horikawa, M., Tsunoda, T., Maoka, T., Matsumoto, S., Simon, J.- C., and Fukatsu, T. (2010). Symbiotic Bacterium Modifies Aphid Body Color. Science 330, 1102–1104.

Venn, A.A., Loram, J.E., and Douglas, A.E. (2008). Photosynthetic symbioses in animals. J Exp Bot 59, 1069–1080.

Vigneron, A., Masson, F., Vallier, A., Balmand, S., Rey, M., Vincent-Monégat, C., Aksoy, E., Aubailly-Giraud, E., Zaidman-Rémy, A., and Heddi, A. (2014). Insects recycle endosymbionts when the benefit is over. Curr. Biol. 24, 2267–2273.

Visick, K.L., and Ruby, E.G. (2006). Vibrio fischeri and its host: it takes two to tango. Current Opinion in Microbiology 9, 632–638.

Walker, T.L., Purton, S., Becker, D.K., and Collet, C. (2005). Microalgae as bioreactors. Plant Cell Rep 24, 629–641.

Warnes, G.R., Bolker, B., Bonebakker, L., Gentleman, R., Huber, W., Liaw, A., Lumley, T., Maechler, M., Magnusson, A., and Moeller, S. (2009). gplots: Various R programming tools for plotting data. R Package Version 2, 1.

Weis, D.S. (1978). Correlation of Infectivity and Concanavalin A Agglutinability of Algae Exsymbiotic from Paramecium bursaria. The Journal of Protozoology 25, 366–370.

Weis, V.M. (2008). Cellular mechanisms of Cnidarian bleaching: stress causes the collapse of symbiosis. Journal of Experimental Biology 211, 3059–3066.

Wendling, C.C., Fabritzek, A.G., and Wegner, K.M. (2017). Population‐specific genotype x genotype x environment interactions in bacterial disease of early life stages of Pacific oyster larvae. Evol Appl 10, 338–347.

Wernegreen, J.J. (2012). Endosymbiosis. Current Biology 22, R555–R561.

Werner, G.D.A., Cornelissen, J.H.C., Cornwell, W.K., Soudzilovskaia, N.A., Kattge, J., West, S.A., and Kiers, E.T. (2018). Symbiont switching and alternative resource acquisition strategies drive mutualism breakdown. PNAS 115, 5229–5234.

Werren, J.H., Baldo, L., and Clark, M.E. (2008). Wolbachia: master manipulators of invertebrate biology. Nature Reviews Microbiology 6, 741–751.

West, S.A., Fisher, R.M., Gardner, A., and Kiers, E.T. (2015). Major evolutionary transitions in individuality. PNAS 112, 10112–10119.

Wichterman, R. (1986). The Biology of Paramecium (Springer US).

Wickham, H. (2016). ggplot2: Elegant Graphics for Data Analysis.

161

Wilkerson, F.P., Kobayashi, D., and Muscatine, L. (1988). Mitotic index and size of symbiotic algae in Caribbean Reef corals. Coral Reefs 7, 29–36.

Wilkinson, H.H., and Parker, M.A. (1996). Symbiotic specialization and the potential for genotypic coexistence in a plant-bacterial mutualism. Oecologia 108, 361–367.

Wilkinson, T.L., and Ishikawa, H. (2000). Injection of essential amino acids substitutes for bacterial supply in aposymbiotic pea aphids (Acyrthosiphon pisum). Entomologia Experimentalis et Applicata 94, 85–91.

Wilson, D.S., and Yoshimura, J. (1994). On the Coexistence of Specialists and Generalists. The American Naturalist 144, 692–707.

Wilson, A.C.C., Ashton, P.D., Calevro, F., Charles, H., Colella, S., Febvay, G., Jander, G., Kushlan, P.F., Macdonald, S.J., Schwartz, J.F., et al. (2010). Genomic insight into the amino acid relations of the pea aphid, Acyrthosiphon pisum, with its symbiotic bacterium Buchnera aphidicola. Insect Molecular Biology 19, 249–258.

Yakovleva, I.M., Baird, A.H., Yamamoto, H.H., Bhagooli, R., Nonaka, M., and Hidaka, M. (2009). Algal symbionts increase oxidative damage and death in coral larvae at high temperatures. Marine Ecology Progress Series 378, 105–112.

Ye, S., Bhattacharjee, M., and Siemann, E. (2019). Thermal Tolerance in Green Hydra: Identifying the Roles of Algal Endosymbionts and Hosts in a Freshwater Holobiont Under Stress. Microb Ecol 77, 537–545.

Yellowlees, D., Rees, T.A.V., and Leggat, W. (2008). Metabolic interactions between algal symbionts and invertebrate hosts. Plant, Cell & Environment 31, 679–694.

Yoshida, T., Hairston, N.G., and Ellner, S.P. (2004). Evolutionary trade–off between defence against grazing and competitive ability in a simple unicellular alga, Chlorella vulgaris. Proceedings of the Royal Society of London. Series B: Biological Sciences 271, 1947–1953.

Zagata, P., Greczek-Stachura, M., Tarcz, S., and Rautian, M. (2016). The Evolutionary Relationships between Endosymbiotic Green Algae of Paramecium bursaria Syngens Originating from Different Geographical Locations. Folia Biologica 64, 47–54.

Zhang, M., Kong, F., Shi, X., Xing, P., and Tan, X. (2007). Differences in Responses to Darkness between Microcystis aeruginosa and Chlorella pyrenoidosa. Journal of Freshwater Ecology 22, 93–99.

Ziesenisz, E., Reisser, W., and Wiessner, W. (1981). Evidence of de novo synthesis of maltose excreted by the endosymbiotic Chlorella from Paramecium bursaria. Planta 153, 481–485.

Zook, D.P. (2002). Prioritizing Symbiosis to Sustain Biodiversity: Are Symbionts Keystone Species? In Symbiosis: Mechanisms and Model Systems, J. Seckbach, ed. (Dordrecht: Springer Netherlands), pp. 3–12.

Zouache, K., Fontaine, A., Vega-Rua, A., Mousson, L., Thiberge, J.-M., Lourenco-De- Oliveira, R., Caro, V., Lambrechts, L., and Failloux, A.-B. (2014). Three-way interactions

162 between mosquito population, viral strain and temperature underlying chikungunya virus transmission potential. Proc. Biol. Sci. 281.

163